File size: 4,327 Bytes
0a56877 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 | ---
license: other
language:
- en
- es
- fr
- de
- it
- pt
- ru
- ar
- hi
- ko
- zh
library_name: transformers
base_model:
- arcee-ai/Trinity-Nano-Preview
base_model_relation: quantized
tags:
- moe
- nvfp4
- modelopt
- blackwell
- vllm
license_link: LICENSE
license_name: openmdw-1.1
---
<div align="center">
<picture>
<img
src="https://cdn-uploads.huggingface.co/production/uploads/6435718aaaef013d1aec3b8b/i-v1KyAMOW_mgVGeic9WJ.png"
alt="Arcee Trinity Nano"
style="max-width: 100%; height: auto;"
>
</picture>
</div>
# Trinity Nano Preview NVFP4
Trinity Nano Preview is a preview of Arcee AI's 6B MoE model with 1B active parameters. It is the small-sized model in our new Trinity family, a series of open-weight models for enterprise and tinkerers alike.
This is a chat tuned model, with a delightful personality and charm we think users will love. We note that this model is pushing the limits of sparsity in small language models with only 800M non-embedding parameters active per token, and as such **may be unstable** in certain use cases, especially in this preview.
This is an *experimental* release, it's fun to talk to but will not be hosted anywhere, so download it and try it out yourself!
***
Trinity Nano Preview is trained on 10T tokens gathered and curated through a key partnership with [Datology](https://www.datologyai.com/), building upon the excellent dataset we used on [AFM-4.5B](https://huggingface.co/arcee-ai/AFM-4.5B) with additional math and code.
Training was performed on a cluster of 512 H200 GPUs powered by [Prime Intellect](https://www.primeintellect.ai/) using HSDP parallelism.
More details, including key architecture decisions, can be found on our blog [here](https://www.arcee.ai/blog/the-trinity-manifesto)
***
**This repository contains the NVFP4 quantized weights of Trinity-Nano-Preview for deployment on NVIDIA Blackwell GPUs.**
## Model Details
* **Model Architecture:** AfmoeForCausalLM
* **Parameters:** 6B, 1B active
* **Experts:** 128 total, 8 active, 1 shared
* **Context length:** 128k
* **Training Tokens:** 10T
* **License:** [OpenMDW-1.1](https://huggingface.co/arcee-ai/Trinity-Nano-Preview#license)
***
<div align="center">
<picture>
<img src="https://cdn-uploads.huggingface.co/production/uploads/6435718aaaef013d1aec3b8b/sSVjGNHfrJKmQ6w8I18ek.png" style="background-color:ghostwhite;padding:5px;" width="17%" alt="Powered by Datology">
</picture>
</div>
## Quantization Details
- **Scheme:** NVFP4 (`nvfp4_mlp_only` — MLP/expert weights only, attention remains BF16)
- **Tool:** [NVIDIA ModelOpt](https://github.com/NVIDIA/Model-Optimizer)
- **Calibration:** 512 samples, seq_length=2048, all-expert calibration enabled
- **KV cache:** Not quantized
## Running with vLLM
Requires [vLLM](https://github.com/vllm-project/vllm) >= 0.18.0. Native FP4 compute requires Blackwell GPUs; older GPUs fall back to Marlin weight decompression automatically.
### Blackwell GPUs (B200/B300/GB300) — Docker (recommended)
```bash
docker run --runtime nvidia --gpus all -p 8000:8000 \
-v ~/.cache/huggingface:/root/.cache/huggingface \
vllm/vllm-openai:v0.18.0-cu130 \
arcee-ai/Trinity-Nano-Preview-NVFP4 \
--trust-remote-code \
--gpu-memory-utilization 0.90 \
--max-model-len 8192
```
### Hopper GPUs (H100/H200) and others
```bash
vllm serve arcee-ai/Trinity-Nano-Preview-NVFP4 \
--trust-remote-code \
--gpu-memory-utilization 0.90 \
--max-model-len 8192 \
--host 0.0.0.0 \
--port 8000
```
**Note (Blackwell pip installs):** If installing vLLM via pip on Blackwell rather than using Docker, native FP4 kernels may produce incorrect output due to package version mismatches. As a workaround, force the Marlin backend:
```bash
export VLLM_NVFP4_GEMM_BACKEND=marlin
vllm serve arcee-ai/Trinity-Nano-Preview-NVFP4 \
--trust-remote-code \
--moe-backend marlin \
--gpu-memory-utilization 0.90 \
--max-model-len 8192 \
--host 0.0.0.0 \
--port 8000
```
Marlin decompresses FP4 weights to BF16 for compute, providing the full memory compression benefit (~3.7× vs BF16) but not native FP4 compute speedup. On Hopper GPUs (H100/H200), Marlin is selected automatically and no extra flags are needed.
## License
Trinity-Nano-Preview-NVFP4 is released under the OpenMDW-1.1 license. |