Add model zoo for the paper's 70 visual tokenizers
Browse filesDocument verified training data and one-epoch schedules, link pretraining projectors and finetuned checkpoints by language backbone, and mark the five unreleased discrete tokenizers.
README.md
CHANGED
|
@@ -21,6 +21,259 @@ The evaluation target is the vision encoder. Runs use different language backbon
|
|
| 21 |
| qwen3 | 978 | 214 | 305.64 |
|
| 22 |
| qwen25 | 1173 | 268 | 326.19 |
|
| 23 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 24 |
## Download
|
| 25 |
|
| 26 |
```python
|
|
|
|
| 21 |
| qwen3 | 978 | 214 | 305.64 |
|
| 22 |
| qwen25 | 1173 | 268 | 326.19 |
|
| 23 |
|
| 24 |
+
## Model Zoo
|
| 25 |
+
|
| 26 |
+
This Model Zoo covers the **70 visual tokenizers used in our paper**: **43 language-supervised**, **22 self-supervised**, and **5 discrete** tokenizers. The tables follow the format of the [LLaVA Model Zoo](https://github.com/haotian-liu/LLaVA/blob/main/docs/MODEL_ZOO.md).
|
| 27 |
+
|
| 28 |
+
The current release provides both pretraining and finetuning checkpoints for all **65 continuous tokenizers** with each of the three main language backbones, plus **2** Qwen3-1.7B-Base runs. The **5 discrete tokenizers** are listed separately with their checkpoint availability. The repository-wide file totals above include all uploaded runs; the catalog below is restricted to the paper's 70 tokenizers.
|
| 29 |
+
|
| 30 |
+
### Training recipe and checkpoint types
|
| 31 |
+
|
| 32 |
+
- **Pretrain Data — LCS-558K:** the image–text alignment dataset from [LLaVA-Pretrain](https://huggingface.co/datasets/liuhaotian/LLaVA-Pretrain), using `blip_laion_cc_sbu_558k.json`. This column describes MLLM projector training, rather than the visual tokenizer's original pretraining corpus.
|
| 33 |
+
- **Finetuning Data — LLaVA-v1.5 mix665k (filtered):** the [LLaVA-v1.5 instruction mixture](https://huggingface.co/datasets/liuhaotian/LLaVA-Instruct-150K/blob/main/llava_v1_5_mix665k.json), using `llava_v1_5_mix665k_drop_ge8kchars.json`. The training configuration removes 395 examples with at least 8,000 characters of conversation text.
|
| 34 |
+
- **`proj-1e`:** train the two-layer GELU MLP projector (`mlp2x_gelu`) for 1 epoch, with the language model and vision encoder frozen.
|
| 35 |
+
- **`full_ft-1e`:** finetune the language model and projector for 1 epoch, with the vision encoder frozen. The released configurations set `mm_tunable_parts` to `mm_mlp_adapter,mm_language_model`.
|
| 36 |
+
- **Download:** `projector` downloads the pretraining `mm_projector.bin`; `finetuned` opens the finetuned checkpoint directory, including model weights, configuration, and language-tokenizer files. The frozen vision encoder must be supplied separately using the matching architecture and weights recorded in `config.json`. Pretraining projector weights alone are not instruction-tuned MLLMs.
|
| 37 |
+
|
| 38 |
+
Tokenizer IDs below match the checkpoint directory names. Schedules are checked against each released run's `trainer_state.json`. Download links point to the main training-stage directories and exclude historical backup directories.
|
| 39 |
+
|
| 40 |
+
### Qwen2.5-1.5B-Instruct
|
| 41 |
+
|
| 42 |
+
Base LLM: [Qwen/Qwen2.5-1.5B-Instruct](https://huggingface.co/Qwen/Qwen2.5-1.5B-Instruct). **65 tokenizers**, each with pretraining and finetuning checkpoints.
|
| 43 |
+
|
| 44 |
+
| Base LLM | Vision Encoder / Tokenizer | Pretrain Data | Pretraining Schedule | Finetuning Data | Finetuning Schedule | Download |
|
| 45 |
+
| --- | --- | --- | --- | --- | --- | --- |
|
| 46 |
+
| Qwen2.5-1.5B-Instruct | PE-Core-G/14 (448)<br>`pe_core_g14_448` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/pe_core_g14_448/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/pe_core_g14_448/finetune) |
|
| 47 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP 2 ViT-G/14 (378)<br>`mc2_g14_378` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc2_g14_378/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc2_g14_378/finetune) |
|
| 48 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP 2 ViT-G/14 (224)<br>`mc2_g14_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc2_g14_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc2_g14_224/finetune) |
|
| 49 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP ViT-G/14 (2.5B, 224)<br>`mc1_g14_224_2.5b` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc1_g14_224_2.5b/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc1_g14_224_2.5b/finetune) |
|
| 50 |
+
| Qwen2.5-1.5B-Instruct | SigLIP2 ViT-G/16 (384)<br>`siglip2_g16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/siglip2_g16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/siglip2_g16_384/finetune) |
|
| 51 |
+
| Qwen2.5-1.5B-Instruct | SigLIP2 ViT-G/16 (256)<br>`siglip2_g16_256` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/siglip2_g16_256/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/siglip2_g16_256/finetune) |
|
| 52 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP 2 ViT-H/14 (378)<br>`mc2_h14_378` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc2_h14_378/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc2_h14_378/finetune) |
|
| 53 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP ViT-H/14 (2.5B, 224)<br>`mc1_h14_224_2.5b` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc1_h14_224_2.5b/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc1_h14_224_2.5b/finetune) |
|
| 54 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP ViT-H/14 (v1.2, 224)<br>`mc1_h14_224_v1.2` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc1_h14_224_v1.2/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc1_h14_224_v1.2/finetune) |
|
| 55 |
+
| Qwen2.5-1.5B-Instruct | OpenAI CLIP ViT-L/14 (224)<br>`clip_openai__l14` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/clip_openai__l14/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/clip_openai__l14/finetune) |
|
| 56 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP 2 ViT-L/14 (224)<br>`mc2_l14_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc2_l14_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc2_l14_224/finetune) |
|
| 57 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP ViT-L/14 (2.5B, 224)<br>`mc1_l14_224_2.5b` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc1_l14_224_2.5b/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc1_l14_224_2.5b/finetune) |
|
| 58 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP ViT-L/14 (400M, 224)<br>`mc1_l14_224_400m` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc1_l14_224_400m/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc1_l14_224_400m/finetune) |
|
| 59 |
+
| Qwen2.5-1.5B-Instruct | SigLIP2 So400m/l14 (384)<br>`siglip2_sm14_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/siglip2_sm14_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/siglip2_sm14_384/finetune) |
|
| 60 |
+
| Qwen2.5-1.5B-Instruct | SigLIP2 So400m/l14 (224)<br>`siglip2_sm14_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/siglip2_sm14_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/siglip2_sm14_224/finetune) |
|
| 61 |
+
| Qwen2.5-1.5B-Instruct | SigLIP2 So400m/l16 (512)<br>`siglip2_sm16_512` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/siglip2_sm16_512/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/siglip2_sm16_512/finetune) |
|
| 62 |
+
| Qwen2.5-1.5B-Instruct | SigLIP2 So400m/l16 (384)<br>`siglip2_sm16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/siglip2_sm16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/siglip2_sm16_384/finetune) |
|
| 63 |
+
| Qwen2.5-1.5B-Instruct | SigLIP2 So400m/l16 (256)<br>`siglip2_sm16_256` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/siglip2_sm16_256/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/siglip2_sm16_256/finetune) |
|
| 64 |
+
| Qwen2.5-1.5B-Instruct | PE-Lang-L/14 (448)<br>`pe_lang_l14_448` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/pe_lang_l14_448/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/pe_lang_l14_448/finetune) |
|
| 65 |
+
| Qwen2.5-1.5B-Instruct | SigLIP2 ViT-L/16 (512)<br>`siglip2_l16_512` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/siglip2_l16_512/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/siglip2_l16_512/finetune) |
|
| 66 |
+
| Qwen2.5-1.5B-Instruct | SigLIP2 ViT-L/16 (384)<br>`siglip2_l16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/siglip2_l16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/siglip2_l16_384/finetune) |
|
| 67 |
+
| Qwen2.5-1.5B-Instruct | SigLIP2 ViT-L/16 (256)<br>`siglip2_l16_256` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/siglip2_l16_256/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/siglip2_l16_256/finetune) |
|
| 68 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP 2 ViT-M/16 (384)<br>`mc2_m16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc2_m16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc2_m16_384/finetune) |
|
| 69 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP 2 ViT-M/16 (224)<br>`mc2_m16_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc2_m16_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc2_m16_224/finetune) |
|
| 70 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP 2 ViT-M/16 (mT5, 224)<br>`mc2_m16_224_mt5` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc2_m16_224_mt5/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc2_m16_224_mt5/finetune) |
|
| 71 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP 2 ViT-B/16 (384)<br>`mc2_b16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc2_b16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc2_b16_384/finetune) |
|
| 72 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP 2 ViT-B/16 (224)<br>`mc2_b16_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc2_b16_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc2_b16_224/finetune) |
|
| 73 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP 2 ViT-B/32 (384)<br>`mc2_b32_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc2_b32_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc2_b32_384/finetune) |
|
| 74 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP 2 ViT-B/32 (224)<br>`mc2_b32_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc2_b32_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc2_b32_224/finetune) |
|
| 75 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP 2 ViT-B/32 (mT5, 224)<br>`mc2_b32_224_mt5` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc2_b32_224_mt5/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc2_b32_224_mt5/finetune) |
|
| 76 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP ViT-B/16 (2.5B, 224)<br>`mc1_b16_224_2.5b` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc1_b16_224_2.5b/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc1_b16_224_2.5b/finetune) |
|
| 77 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP ViT-B/32 (2.5B, 224)<br>`mc1_b32_224_2.5b` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc1_b32_224_2.5b/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc1_b32_224_2.5b/finetune) |
|
| 78 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP ViT-B/16 (400M, 224)<br>`mc1_b16_224_400m` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc1_b16_224_400m/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc1_b16_224_400m/finetune) |
|
| 79 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP ViT-B/32 (400M, 224)<br>`mc1_b32_224_400m` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc1_b32_224_400m/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc1_b32_224_400m/finetune) |
|
| 80 |
+
| Qwen2.5-1.5B-Instruct | PE-Core-B/16 (224)<br>`pe_core_b16_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/pe_core_b16_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/pe_core_b16_224/finetune) |
|
| 81 |
+
| Qwen2.5-1.5B-Instruct | SigLIP2 ViT-B/16 (512)<br>`siglip2_b16_512` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/siglip2_b16_512/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/siglip2_b16_512/finetune) |
|
| 82 |
+
| Qwen2.5-1.5B-Instruct | SigLIP2 ViT-B/16 (384)<br>`siglip2_b16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/siglip2_b16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/siglip2_b16_384/finetune) |
|
| 83 |
+
| Qwen2.5-1.5B-Instruct | SigLIP2 ViT-B/16 (256)<br>`siglip2_b16_256` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/siglip2_b16_256/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/siglip2_b16_256/finetune) |
|
| 84 |
+
| Qwen2.5-1.5B-Instruct | SigLIP2 ViT-B/16 (224)<br>`siglip2_b16_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/siglip2_b16_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/siglip2_b16_224/finetune) |
|
| 85 |
+
| Qwen2.5-1.5B-Instruct | SigLIP2 ViT-B/32 (256)<br>`siglip2_b32_256` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/siglip2_b32_256/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/siglip2_b32_256/finetune) |
|
| 86 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP 2 ViT-S/16 (384)<br>`mc2_s16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc2_s16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc2_s16_384/finetune) |
|
| 87 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP 2 ViT-S/16 (224)<br>`mc2_s16_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc2_s16_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc2_s16_224/finetune) |
|
| 88 |
+
| Qwen2.5-1.5B-Instruct | MetaCLIP 2 ViT-S/16 (mT5, 224)<br>`mc2_s16_224_mt5` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/mc2_s16_224_mt5/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/mc2_s16_224_mt5/finetune) |
|
| 89 |
+
| Qwen2.5-1.5B-Instruct | Web-SSL MAE 3B (224)<br>`webssl_mae3b_full2b_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/webssl_mae3b_full2b_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/webssl_mae3b_full2b_224/finetune) |
|
| 90 |
+
| Qwen2.5-1.5B-Instruct | DINOv2 ViT-G/14<br>`dinov2_giant` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/dinov2_giant/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/dinov2_giant/finetune) |
|
| 91 |
+
| Qwen2.5-1.5B-Instruct | Web-SSL DINO 1B (224)<br>`webssl_dino1b_full2b_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/webssl_dino1b_full2b_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/webssl_dino1b_full2b_224/finetune) |
|
| 92 |
+
| Qwen2.5-1.5B-Instruct | Web-SSL MAE 1B (224)<br>`webssl_mae1b_full2b_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/webssl_mae1b_full2b_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/webssl_mae1b_full2b_224/finetune) |
|
| 93 |
+
| Qwen2.5-1.5B-Instruct | Pixio ViT-H/16<br>`pixio_vith16` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/pixio_vith16/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/pixio_vith16/finetune) |
|
| 94 |
+
| Qwen2.5-1.5B-Instruct | I-JEPA ViT-H/14<br>`ijepa_vith14` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/ijepa_vith14/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/ijepa_vith14/finetune) |
|
| 95 |
+
| Qwen2.5-1.5B-Instruct | DINOv2 ViT-L/14<br>`dinov2_large` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/dinov2_large/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/dinov2_large/finetune) |
|
| 96 |
+
| Qwen2.5-1.5B-Instruct | DINOv3 ViT-L/16<br>`dinov3_vitl16` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/dinov3_vitl16/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/dinov3_vitl16/finetune) |
|
| 97 |
+
| Qwen2.5-1.5B-Instruct | Pixio ViT-L/16<br>`pixio_vitl16` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/pixio_vitl16/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/pixio_vitl16/finetune) |
|
| 98 |
+
| Qwen2.5-1.5B-Instruct | RAEv2 DINOv3-L (k=7)<br>`raev2_dinov3l_k7` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/raev2_dinov3l_k7/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/raev2_dinov3l_k7/finetune) |
|
| 99 |
+
| Qwen2.5-1.5B-Instruct | Web-SSL MAE 300M (224)<br>`webssl_mae300m_full2b_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/webssl_mae300m_full2b_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/webssl_mae300m_full2b_224/finetune) |
|
| 100 |
+
| Qwen2.5-1.5B-Instruct | EUPE ConvNeXt-B<br>`eupe_convnext_b` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/eupe_convnext_b/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/eupe_convnext_b/finetune) |
|
| 101 |
+
| Qwen2.5-1.5B-Instruct | DINOv2 ViT-B/14<br>`dinov2_base` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/dinov2_base/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/dinov2_base/finetune) |
|
| 102 |
+
| Qwen2.5-1.5B-Instruct | DINO ViT-B/8<br>`dino_vitb8` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/dino_vitb8/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/dino_vitb8/finetune) |
|
| 103 |
+
| Qwen2.5-1.5B-Instruct | DINO ViT-B/16<br>`dino_vitb16` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/dino_vitb16/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/dino_vitb16/finetune) |
|
| 104 |
+
| Qwen2.5-1.5B-Instruct | EUPE ViT-B<br>`eupe_vit_b` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/eupe_vit_b/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/eupe_vit_b/finetune) |
|
| 105 |
+
| Qwen2.5-1.5B-Instruct | Pixio ViT-B/16<br>`pixio_vitb16` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/pixio_vitb16/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/pixio_vitb16/finetune) |
|
| 106 |
+
| Qwen2.5-1.5B-Instruct | DINOv2 ViT-S/14<br>`dinov2_small` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/dinov2_small/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/dinov2_small/finetune) |
|
| 107 |
+
| Qwen2.5-1.5B-Instruct | DINO ViT-S/16<br>`dino_vits16` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/dino_vits16/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/dino_vits16/finetune) |
|
| 108 |
+
| Qwen2.5-1.5B-Instruct | EUPE ViT-S<br>`eupe_vit_s` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/eupe_vit_s/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/eupe_vit_s/finetune) |
|
| 109 |
+
| Qwen2.5-1.5B-Instruct | DINO ViT-S/8<br>`dino_vits8` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/dino_vits8/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/dino_vits8/finetune) |
|
| 110 |
+
| Qwen2.5-1.5B-Instruct | EUPE ViT-T<br>`eupe_vit_t` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen25/eupe_vit_t/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen25/eupe_vit_t/finetune) |
|
| 111 |
+
|
| 112 |
+
### Qwen3-1.7B
|
| 113 |
+
|
| 114 |
+
Base LLM: [Qwen/Qwen3-1.7B](https://huggingface.co/Qwen/Qwen3-1.7B). **65 tokenizers**, each with pretraining and finetuning checkpoints.
|
| 115 |
+
|
| 116 |
+
| Base LLM | Vision Encoder / Tokenizer | Pretrain Data | Pretraining Schedule | Finetuning Data | Finetuning Schedule | Download |
|
| 117 |
+
| --- | --- | --- | --- | --- | --- | --- |
|
| 118 |
+
| Qwen3-1.7B | PE-Core-G/14 (448)<br>`pe_core_g14_448` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/pe_core_g14_448/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/pe_core_g14_448/finetune) |
|
| 119 |
+
| Qwen3-1.7B | MetaCLIP 2 ViT-G/14 (378)<br>`mc2_g14_378` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc2_g14_378/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc2_g14_378/finetune) |
|
| 120 |
+
| Qwen3-1.7B | MetaCLIP 2 ViT-G/14 (224)<br>`mc2_g14_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc2_g14_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc2_g14_224/finetune) |
|
| 121 |
+
| Qwen3-1.7B | MetaCLIP ViT-G/14 (2.5B, 224)<br>`mc1_g14_224_2.5b` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc1_g14_224_2.5b/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc1_g14_224_2.5b/finetune) |
|
| 122 |
+
| Qwen3-1.7B | SigLIP2 ViT-G/16 (384)<br>`siglip2_g16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/siglip2_g16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/siglip2_g16_384/finetune) |
|
| 123 |
+
| Qwen3-1.7B | SigLIP2 ViT-G/16 (256)<br>`siglip2_g16_256` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/siglip2_g16_256/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/siglip2_g16_256/finetune) |
|
| 124 |
+
| Qwen3-1.7B | MetaCLIP 2 ViT-H/14 (378)<br>`mc2_h14_378` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc2_h14_378/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc2_h14_378/finetune) |
|
| 125 |
+
| Qwen3-1.7B | MetaCLIP ViT-H/14 (2.5B, 224)<br>`mc1_h14_224_2.5b` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc1_h14_224_2.5b/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc1_h14_224_2.5b/finetune) |
|
| 126 |
+
| Qwen3-1.7B | MetaCLIP ViT-H/14 (v1.2, 224)<br>`mc1_h14_224_v1.2` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc1_h14_224_v1.2/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc1_h14_224_v1.2/finetune) |
|
| 127 |
+
| Qwen3-1.7B | OpenAI CLIP ViT-L/14 (224)<br>`clip_openai__l14` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/clip_openai__l14/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/clip_openai__l14/finetune) |
|
| 128 |
+
| Qwen3-1.7B | MetaCLIP 2 ViT-L/14 (224)<br>`mc2_l14_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc2_l14_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc2_l14_224/finetune) |
|
| 129 |
+
| Qwen3-1.7B | MetaCLIP ViT-L/14 (2.5B, 224)<br>`mc1_l14_224_2.5b` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc1_l14_224_2.5b/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc1_l14_224_2.5b/finetune) |
|
| 130 |
+
| Qwen3-1.7B | MetaCLIP ViT-L/14 (400M, 224)<br>`mc1_l14_224_400m` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc1_l14_224_400m/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc1_l14_224_400m/finetune) |
|
| 131 |
+
| Qwen3-1.7B | SigLIP2 So400m/l14 (384)<br>`siglip2_sm14_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/siglip2_sm14_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/siglip2_sm14_384/finetune) |
|
| 132 |
+
| Qwen3-1.7B | SigLIP2 So400m/l14 (224)<br>`siglip2_sm14_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/siglip2_sm14_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/siglip2_sm14_224/finetune) |
|
| 133 |
+
| Qwen3-1.7B | SigLIP2 So400m/l16 (512)<br>`siglip2_sm16_512` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/siglip2_sm16_512/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/siglip2_sm16_512/finetune) |
|
| 134 |
+
| Qwen3-1.7B | SigLIP2 So400m/l16 (384)<br>`siglip2_sm16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/siglip2_sm16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/siglip2_sm16_384/finetune) |
|
| 135 |
+
| Qwen3-1.7B | SigLIP2 So400m/l16 (256)<br>`siglip2_sm16_256` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/siglip2_sm16_256/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/siglip2_sm16_256/finetune) |
|
| 136 |
+
| Qwen3-1.7B | PE-Lang-L/14 (448)<br>`pe_lang_l14_448` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/pe_lang_l14_448/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/pe_lang_l14_448/finetune) |
|
| 137 |
+
| Qwen3-1.7B | SigLIP2 ViT-L/16 (512)<br>`siglip2_l16_512` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/siglip2_l16_512/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/siglip2_l16_512/finetune) |
|
| 138 |
+
| Qwen3-1.7B | SigLIP2 ViT-L/16 (384)<br>`siglip2_l16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/siglip2_l16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/siglip2_l16_384/finetune) |
|
| 139 |
+
| Qwen3-1.7B | SigLIP2 ViT-L/16 (256)<br>`siglip2_l16_256` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/siglip2_l16_256/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/siglip2_l16_256/finetune) |
|
| 140 |
+
| Qwen3-1.7B | MetaCLIP 2 ViT-M/16 (384)<br>`mc2_m16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc2_m16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc2_m16_384/finetune) |
|
| 141 |
+
| Qwen3-1.7B | MetaCLIP 2 ViT-M/16 (224)<br>`mc2_m16_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc2_m16_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc2_m16_224/finetune) |
|
| 142 |
+
| Qwen3-1.7B | MetaCLIP 2 ViT-M/16 (mT5, 224)<br>`mc2_m16_224_mt5` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc2_m16_224_mt5/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc2_m16_224_mt5/finetune) |
|
| 143 |
+
| Qwen3-1.7B | MetaCLIP 2 ViT-B/16 (384)<br>`mc2_b16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc2_b16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc2_b16_384/finetune) |
|
| 144 |
+
| Qwen3-1.7B | MetaCLIP 2 ViT-B/16 (224)<br>`mc2_b16_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc2_b16_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc2_b16_224/finetune) |
|
| 145 |
+
| Qwen3-1.7B | MetaCLIP 2 ViT-B/32 (384)<br>`mc2_b32_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc2_b32_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc2_b32_384/finetune) |
|
| 146 |
+
| Qwen3-1.7B | MetaCLIP 2 ViT-B/32 (224)<br>`mc2_b32_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc2_b32_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc2_b32_224/finetune) |
|
| 147 |
+
| Qwen3-1.7B | MetaCLIP 2 ViT-B/32 (mT5, 224)<br>`mc2_b32_224_mt5` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc2_b32_224_mt5/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc2_b32_224_mt5/finetune) |
|
| 148 |
+
| Qwen3-1.7B | MetaCLIP ViT-B/16 (2.5B, 224)<br>`mc1_b16_224_2.5b` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc1_b16_224_2.5b/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc1_b16_224_2.5b/finetune) |
|
| 149 |
+
| Qwen3-1.7B | MetaCLIP ViT-B/32 (2.5B, 224)<br>`mc1_b32_224_2.5b` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc1_b32_224_2.5b/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc1_b32_224_2.5b/finetune) |
|
| 150 |
+
| Qwen3-1.7B | MetaCLIP ViT-B/16 (400M, 224)<br>`mc1_b16_224_400m` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc1_b16_224_400m/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc1_b16_224_400m/finetune) |
|
| 151 |
+
| Qwen3-1.7B | MetaCLIP ViT-B/32 (400M, 224)<br>`mc1_b32_224_400m` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc1_b32_224_400m/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc1_b32_224_400m/finetune) |
|
| 152 |
+
| Qwen3-1.7B | PE-Core-B/16 (224)<br>`pe_core_b16_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/pe_core_b16_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/pe_core_b16_224/finetune) |
|
| 153 |
+
| Qwen3-1.7B | SigLIP2 ViT-B/16 (512)<br>`siglip2_b16_512` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/siglip2_b16_512/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/siglip2_b16_512/finetune) |
|
| 154 |
+
| Qwen3-1.7B | SigLIP2 ViT-B/16 (384)<br>`siglip2_b16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/siglip2_b16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/siglip2_b16_384/finetune) |
|
| 155 |
+
| Qwen3-1.7B | SigLIP2 ViT-B/16 (256)<br>`siglip2_b16_256` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/siglip2_b16_256/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/siglip2_b16_256/finetune) |
|
| 156 |
+
| Qwen3-1.7B | SigLIP2 ViT-B/16 (224)<br>`siglip2_b16_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/siglip2_b16_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/siglip2_b16_224/finetune) |
|
| 157 |
+
| Qwen3-1.7B | SigLIP2 ViT-B/32 (256)<br>`siglip2_b32_256` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/siglip2_b32_256/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/siglip2_b32_256/finetune) |
|
| 158 |
+
| Qwen3-1.7B | MetaCLIP 2 ViT-S/16 (384)<br>`mc2_s16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc2_s16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc2_s16_384/finetune) |
|
| 159 |
+
| Qwen3-1.7B | MetaCLIP 2 ViT-S/16 (224)<br>`mc2_s16_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc2_s16_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc2_s16_224/finetune) |
|
| 160 |
+
| Qwen3-1.7B | MetaCLIP 2 ViT-S/16 (mT5, 224)<br>`mc2_s16_224_mt5` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/mc2_s16_224_mt5/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/mc2_s16_224_mt5/finetune) |
|
| 161 |
+
| Qwen3-1.7B | Web-SSL MAE 3B (224)<br>`webssl_mae3b_full2b_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/webssl_mae3b_full2b_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/webssl_mae3b_full2b_224/finetune) |
|
| 162 |
+
| Qwen3-1.7B | DINOv2 ViT-G/14<br>`dinov2_giant` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/dinov2_giant/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/dinov2_giant/finetune) |
|
| 163 |
+
| Qwen3-1.7B | Web-SSL DINO 1B (224)<br>`webssl_dino1b_full2b_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/webssl_dino1b_full2b_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/webssl_dino1b_full2b_224/finetune) |
|
| 164 |
+
| Qwen3-1.7B | Web-SSL MAE 1B (224)<br>`webssl_mae1b_full2b_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/webssl_mae1b_full2b_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/webssl_mae1b_full2b_224/finetune) |
|
| 165 |
+
| Qwen3-1.7B | Pixio ViT-H/16<br>`pixio_vith16` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/pixio_vith16/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/pixio_vith16/finetune) |
|
| 166 |
+
| Qwen3-1.7B | I-JEPA ViT-H/14<br>`ijepa_vith14` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/ijepa_vith14/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/ijepa_vith14/finetune) |
|
| 167 |
+
| Qwen3-1.7B | DINOv2 ViT-L/14<br>`dinov2_large` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/dinov2_large/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/dinov2_large/finetune) |
|
| 168 |
+
| Qwen3-1.7B | DINOv3 ViT-L/16<br>`dinov3_vitl16` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/dinov3_vitl16/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/dinov3_vitl16/finetune) |
|
| 169 |
+
| Qwen3-1.7B | Pixio ViT-L/16<br>`pixio_vitl16` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/pixio_vitl16/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/pixio_vitl16/finetune) |
|
| 170 |
+
| Qwen3-1.7B | RAEv2 DINOv3-L (k=7)<br>`raev2_dinov3l_k7` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/raev2_dinov3l_k7/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/raev2_dinov3l_k7/finetune) |
|
| 171 |
+
| Qwen3-1.7B | Web-SSL MAE 300M (224)<br>`webssl_mae300m_full2b_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/webssl_mae300m_full2b_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/webssl_mae300m_full2b_224/finetune) |
|
| 172 |
+
| Qwen3-1.7B | EUPE ConvNeXt-B<br>`eupe_convnext_b` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/eupe_convnext_b/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/eupe_convnext_b/finetune) |
|
| 173 |
+
| Qwen3-1.7B | DINOv2 ViT-B/14<br>`dinov2_base` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/dinov2_base/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/dinov2_base/finetune) |
|
| 174 |
+
| Qwen3-1.7B | DINO ViT-B/8<br>`dino_vitb8` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/dino_vitb8/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/dino_vitb8/finetune) |
|
| 175 |
+
| Qwen3-1.7B | DINO ViT-B/16<br>`dino_vitb16` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/dino_vitb16/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/dino_vitb16/finetune) |
|
| 176 |
+
| Qwen3-1.7B | EUPE ViT-B<br>`eupe_vit_b` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/eupe_vit_b/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/eupe_vit_b/finetune) |
|
| 177 |
+
| Qwen3-1.7B | Pixio ViT-B/16<br>`pixio_vitb16` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/pixio_vitb16/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/pixio_vitb16/finetune) |
|
| 178 |
+
| Qwen3-1.7B | DINOv2 ViT-S/14<br>`dinov2_small` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/dinov2_small/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/dinov2_small/finetune) |
|
| 179 |
+
| Qwen3-1.7B | DINO ViT-S/16<br>`dino_vits16` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/dino_vits16/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/dino_vits16/finetune) |
|
| 180 |
+
| Qwen3-1.7B | EUPE ViT-S<br>`eupe_vit_s` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/eupe_vit_s/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/eupe_vit_s/finetune) |
|
| 181 |
+
| Qwen3-1.7B | DINO ViT-S/8<br>`dino_vits8` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/dino_vits8/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/dino_vits8/finetune) |
|
| 182 |
+
| Qwen3-1.7B | EUPE ViT-T<br>`eupe_vit_t` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3/eupe_vit_t/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3/eupe_vit_t/finetune) |
|
| 183 |
+
|
| 184 |
+
### SmolLM2-1.7B-Instruct
|
| 185 |
+
|
| 186 |
+
Base LLM: [HuggingFaceTB/SmolLM2-1.7B-Instruct](https://huggingface.co/HuggingFaceTB/SmolLM2-1.7B-Instruct). **65 tokenizers**, each with pretraining and finetuning checkpoints.
|
| 187 |
+
|
| 188 |
+
| Base LLM | Vision Encoder / Tokenizer | Pretrain Data | Pretraining Schedule | Finetuning Data | Finetuning Schedule | Download |
|
| 189 |
+
| --- | --- | --- | --- | --- | --- | --- |
|
| 190 |
+
| SmolLM2-1.7B-Instruct | PE-Core-G/14 (448)<br>`pe_core_g14_448` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/pe_core_g14_448/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/pe_core_g14_448/finetune) |
|
| 191 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP 2 ViT-G/14 (378)<br>`mc2_g14_378` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc2_g14_378/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc2_g14_378/finetune) |
|
| 192 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP 2 ViT-G/14 (224)<br>`mc2_g14_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc2_g14_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc2_g14_224/finetune) |
|
| 193 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP ViT-G/14 (2.5B, 224)<br>`mc1_g14_224_2.5b` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc1_g14_224_2.5b/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc1_g14_224_2.5b/finetune) |
|
| 194 |
+
| SmolLM2-1.7B-Instruct | SigLIP2 ViT-G/16 (384)<br>`siglip2_g16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/siglip2_g16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/siglip2_g16_384/finetune) |
|
| 195 |
+
| SmolLM2-1.7B-Instruct | SigLIP2 ViT-G/16 (256)<br>`siglip2_g16_256` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/siglip2_g16_256/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/siglip2_g16_256/finetune) |
|
| 196 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP 2 ViT-H/14 (378)<br>`mc2_h14_378` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc2_h14_378/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc2_h14_378/finetune) |
|
| 197 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP ViT-H/14 (2.5B, 224)<br>`mc1_h14_224_2.5b` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc1_h14_224_2.5b/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc1_h14_224_2.5b/finetune) |
|
| 198 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP ViT-H/14 (v1.2, 224)<br>`mc1_h14_224_v1.2` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc1_h14_224_v1.2/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc1_h14_224_v1.2/finetune) |
|
| 199 |
+
| SmolLM2-1.7B-Instruct | OpenAI CLIP ViT-L/14 (224)<br>`clip_openai__l14` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/clip_openai__l14/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/clip_openai__l14/finetune) |
|
| 200 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP 2 ViT-L/14 (224)<br>`mc2_l14_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc2_l14_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc2_l14_224/finetune) |
|
| 201 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP ViT-L/14 (2.5B, 224)<br>`mc1_l14_224_2.5b` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc1_l14_224_2.5b/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc1_l14_224_2.5b/finetune) |
|
| 202 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP ViT-L/14 (400M, 224)<br>`mc1_l14_224_400m` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc1_l14_224_400m/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc1_l14_224_400m/finetune) |
|
| 203 |
+
| SmolLM2-1.7B-Instruct | SigLIP2 So400m/l14 (384)<br>`siglip2_sm14_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/siglip2_sm14_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/siglip2_sm14_384/finetune) |
|
| 204 |
+
| SmolLM2-1.7B-Instruct | SigLIP2 So400m/l14 (224)<br>`siglip2_sm14_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/siglip2_sm14_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/siglip2_sm14_224/finetune) |
|
| 205 |
+
| SmolLM2-1.7B-Instruct | SigLIP2 So400m/l16 (512)<br>`siglip2_sm16_512` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/siglip2_sm16_512/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/siglip2_sm16_512/finetune) |
|
| 206 |
+
| SmolLM2-1.7B-Instruct | SigLIP2 So400m/l16 (384)<br>`siglip2_sm16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/siglip2_sm16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/siglip2_sm16_384/finetune) |
|
| 207 |
+
| SmolLM2-1.7B-Instruct | SigLIP2 So400m/l16 (256)<br>`siglip2_sm16_256` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/siglip2_sm16_256/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/siglip2_sm16_256/finetune) |
|
| 208 |
+
| SmolLM2-1.7B-Instruct | PE-Lang-L/14 (448)<br>`pe_lang_l14_448` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/pe_lang_l14_448/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/pe_lang_l14_448/finetune) |
|
| 209 |
+
| SmolLM2-1.7B-Instruct | SigLIP2 ViT-L/16 (512)<br>`siglip2_l16_512` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/siglip2_l16_512/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/siglip2_l16_512/finetune) |
|
| 210 |
+
| SmolLM2-1.7B-Instruct | SigLIP2 ViT-L/16 (384)<br>`siglip2_l16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/siglip2_l16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/siglip2_l16_384/finetune) |
|
| 211 |
+
| SmolLM2-1.7B-Instruct | SigLIP2 ViT-L/16 (256)<br>`siglip2_l16_256` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/siglip2_l16_256/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/siglip2_l16_256/finetune) |
|
| 212 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP 2 ViT-M/16 (384)<br>`mc2_m16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc2_m16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc2_m16_384/finetune) |
|
| 213 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP 2 ViT-M/16 (224)<br>`mc2_m16_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc2_m16_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc2_m16_224/finetune) |
|
| 214 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP 2 ViT-M/16 (mT5, 224)<br>`mc2_m16_224_mt5` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc2_m16_224_mt5/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc2_m16_224_mt5/finetune) |
|
| 215 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP 2 ViT-B/16 (384)<br>`mc2_b16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc2_b16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc2_b16_384/finetune) |
|
| 216 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP 2 ViT-B/16 (224)<br>`mc2_b16_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc2_b16_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc2_b16_224/finetune) |
|
| 217 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP 2 ViT-B/32 (384)<br>`mc2_b32_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc2_b32_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc2_b32_384/finetune) |
|
| 218 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP 2 ViT-B/32 (224)<br>`mc2_b32_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc2_b32_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc2_b32_224/finetune) |
|
| 219 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP 2 ViT-B/32 (mT5, 224)<br>`mc2_b32_224_mt5` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc2_b32_224_mt5/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc2_b32_224_mt5/finetune) |
|
| 220 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP ViT-B/16 (2.5B, 224)<br>`mc1_b16_224_2.5b` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc1_b16_224_2.5b/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc1_b16_224_2.5b/finetune) |
|
| 221 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP ViT-B/32 (2.5B, 224)<br>`mc1_b32_224_2.5b` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc1_b32_224_2.5b/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc1_b32_224_2.5b/finetune) |
|
| 222 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP ViT-B/16 (400M, 224)<br>`mc1_b16_224_400m` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc1_b16_224_400m/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc1_b16_224_400m/finetune) |
|
| 223 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP ViT-B/32 (400M, 224)<br>`mc1_b32_224_400m` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc1_b32_224_400m/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc1_b32_224_400m/finetune) |
|
| 224 |
+
| SmolLM2-1.7B-Instruct | PE-Core-B/16 (224)<br>`pe_core_b16_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/pe_core_b16_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/pe_core_b16_224/finetune) |
|
| 225 |
+
| SmolLM2-1.7B-Instruct | SigLIP2 ViT-B/16 (512)<br>`siglip2_b16_512` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/siglip2_b16_512/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/siglip2_b16_512/finetune) |
|
| 226 |
+
| SmolLM2-1.7B-Instruct | SigLIP2 ViT-B/16 (384)<br>`siglip2_b16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/siglip2_b16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/siglip2_b16_384/finetune) |
|
| 227 |
+
| SmolLM2-1.7B-Instruct | SigLIP2 ViT-B/16 (256)<br>`siglip2_b16_256` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/siglip2_b16_256/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/siglip2_b16_256/finetune) |
|
| 228 |
+
| SmolLM2-1.7B-Instruct | SigLIP2 ViT-B/16 (224)<br>`siglip2_b16_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/siglip2_b16_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/siglip2_b16_224/finetune) |
|
| 229 |
+
| SmolLM2-1.7B-Instruct | SigLIP2 ViT-B/32 (256)<br>`siglip2_b32_256` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/siglip2_b32_256/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/siglip2_b32_256/finetune) |
|
| 230 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP 2 ViT-S/16 (384)<br>`mc2_s16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc2_s16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc2_s16_384/finetune) |
|
| 231 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP 2 ViT-S/16 (224)<br>`mc2_s16_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc2_s16_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc2_s16_224/finetune) |
|
| 232 |
+
| SmolLM2-1.7B-Instruct | MetaCLIP 2 ViT-S/16 (mT5, 224)<br>`mc2_s16_224_mt5` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/mc2_s16_224_mt5/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/mc2_s16_224_mt5/finetune) |
|
| 233 |
+
| SmolLM2-1.7B-Instruct | Web-SSL MAE 3B (224)<br>`webssl_mae3b_full2b_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/webssl_mae3b_full2b_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/webssl_mae3b_full2b_224/finetune) |
|
| 234 |
+
| SmolLM2-1.7B-Instruct | DINOv2 ViT-G/14<br>`dinov2_giant` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/dinov2_giant/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/dinov2_giant/finetune) |
|
| 235 |
+
| SmolLM2-1.7B-Instruct | Web-SSL DINO 1B (224)<br>`webssl_dino1b_full2b_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/webssl_dino1b_full2b_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/webssl_dino1b_full2b_224/finetune) |
|
| 236 |
+
| SmolLM2-1.7B-Instruct | Web-SSL MAE 1B (224)<br>`webssl_mae1b_full2b_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/webssl_mae1b_full2b_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/webssl_mae1b_full2b_224/finetune) |
|
| 237 |
+
| SmolLM2-1.7B-Instruct | Pixio ViT-H/16<br>`pixio_vith16` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/pixio_vith16/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/pixio_vith16/finetune) |
|
| 238 |
+
| SmolLM2-1.7B-Instruct | I-JEPA ViT-H/14<br>`ijepa_vith14` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/ijepa_vith14/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/ijepa_vith14/finetune) |
|
| 239 |
+
| SmolLM2-1.7B-Instruct | DINOv2 ViT-L/14<br>`dinov2_large` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/dinov2_large/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/dinov2_large/finetune) |
|
| 240 |
+
| SmolLM2-1.7B-Instruct | DINOv3 ViT-L/16<br>`dinov3_vitl16` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/dinov3_vitl16/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/dinov3_vitl16/finetune) |
|
| 241 |
+
| SmolLM2-1.7B-Instruct | Pixio ViT-L/16<br>`pixio_vitl16` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/pixio_vitl16/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/pixio_vitl16/finetune) |
|
| 242 |
+
| SmolLM2-1.7B-Instruct | RAEv2 DINOv3-L (k=7)<br>`raev2_dinov3l_k7` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/raev2_dinov3l_k7/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/raev2_dinov3l_k7/finetune) |
|
| 243 |
+
| SmolLM2-1.7B-Instruct | Web-SSL MAE 300M (224)<br>`webssl_mae300m_full2b_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/webssl_mae300m_full2b_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/webssl_mae300m_full2b_224/finetune) |
|
| 244 |
+
| SmolLM2-1.7B-Instruct | EUPE ConvNeXt-B<br>`eupe_convnext_b` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/eupe_convnext_b/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/eupe_convnext_b/finetune) |
|
| 245 |
+
| SmolLM2-1.7B-Instruct | DINOv2 ViT-B/14<br>`dinov2_base` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/dinov2_base/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/dinov2_base/finetune) |
|
| 246 |
+
| SmolLM2-1.7B-Instruct | DINO ViT-B/8<br>`dino_vitb8` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/dino_vitb8/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/dino_vitb8/finetune) |
|
| 247 |
+
| SmolLM2-1.7B-Instruct | DINO ViT-B/16<br>`dino_vitb16` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/dino_vitb16/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/dino_vitb16/finetune) |
|
| 248 |
+
| SmolLM2-1.7B-Instruct | EUPE ViT-B<br>`eupe_vit_b` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/eupe_vit_b/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/eupe_vit_b/finetune) |
|
| 249 |
+
| SmolLM2-1.7B-Instruct | Pixio ViT-B/16<br>`pixio_vitb16` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/pixio_vitb16/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/pixio_vitb16/finetune) |
|
| 250 |
+
| SmolLM2-1.7B-Instruct | DINOv2 ViT-S/14<br>`dinov2_small` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/dinov2_small/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/dinov2_small/finetune) |
|
| 251 |
+
| SmolLM2-1.7B-Instruct | DINO ViT-S/16<br>`dino_vits16` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/dino_vits16/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/dino_vits16/finetune) |
|
| 252 |
+
| SmolLM2-1.7B-Instruct | EUPE ViT-S<br>`eupe_vit_s` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/eupe_vit_s/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/eupe_vit_s/finetune) |
|
| 253 |
+
| SmolLM2-1.7B-Instruct | DINO ViT-S/8<br>`dino_vits8` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/dino_vits8/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/dino_vits8/finetune) |
|
| 254 |
+
| SmolLM2-1.7B-Instruct | EUPE ViT-T<br>`eupe_vit_t` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/smollm2/eupe_vit_t/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/smollm2/eupe_vit_t/finetune) |
|
| 255 |
+
|
| 256 |
+
### Qwen3-1.7B-Base
|
| 257 |
+
|
| 258 |
+
Base LLM: [Qwen/Qwen3-1.7B-Base](https://huggingface.co/Qwen/Qwen3-1.7B-Base). **2 tokenizers** from the same paper inventory have released checkpoints with this additional backbone.
|
| 259 |
+
|
| 260 |
+
| Base LLM | Vision Encoder / Tokenizer | Pretrain Data | Pretraining Schedule | Finetuning Data | Finetuning Schedule | Download |
|
| 261 |
+
| --- | --- | --- | --- | --- | --- | --- |
|
| 262 |
+
| Qwen3-1.7B-Base | SigLIP2 ViT-G/16 (384)<br>`siglip2_g16_384` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3base/siglip2_g16_384/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3base/siglip2_g16_384/finetune) |
|
| 263 |
+
| Qwen3-1.7B-Base | MetaCLIP 2 ViT-L/14 (224)<br>`mc2_l14_224` | LCS-558K | proj-1e | LLaVA-v1.5 mix665k (filtered) | full_ft-1e | [projector](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/resolve/main/continuous/qwen3base/mc2_l14_224/pretrain/mm_projector.bin) / [finetuned](https://huggingface.co/336labs/VisionEncoder-MLLM-Eval/tree/main/continuous/qwen3base/mc2_l14_224/finetune) |
|
| 264 |
+
|
| 265 |
+
### Discrete tokenizers in the paper
|
| 266 |
+
|
| 267 |
+
These five tokenizers complete the paper's 70-tokenizer inventory. Their downstream MLLM checkpoints are **not available in this repository**, so no checkpoint download links or released training schedules are listed.
|
| 268 |
+
|
| 269 |
+
| Vision Tokenizer | Tokenizer ID | Checkpoint Availability |
|
| 270 |
+
| --- | --- | --- |
|
| 271 |
+
| UniTok-Attn (256) | `unitok_attn` | Not released in this repository |
|
| 272 |
+
| UniAR-BSQ | `uniar_bsq` | Not released in this repository |
|
| 273 |
+
| TokLIP-L (384) | `toklip_l_384` | Not released in this repository |
|
| 274 |
+
| TokLIP-S (256) | `toklip_s_256` | Not released in this repository |
|
| 275 |
+
| VILA-U (256) | `vilau_256` | Not released in this repository |
|
| 276 |
+
|
| 277 |
## Download
|
| 278 |
|
| 279 |
```python
|