someorgtoo (Some Org)

posted an update about 18 hours ago

Post

213

We compared 1-bit Kimi K3 to Claude Opus 5 and GPT 5.6. 🤯

We gave 4 models the same prompt: Create a glass aquarium whose side panel develops a visible crack and then bursts...

1-bit Kimi K3 GGUF ran locally on 4x B200s at 36 tok/s.

GGUF: unsloth/Kimi-K3-GGUF
GitHub repo: https://github.com/unslothai/unsloth

1 reply

·

danielhanchen

posted an update 3 days ago

Post

3533

Kimi K3 can now be run locally! ✨

The 1-bit model retains ~78.9% accuracy after we shrunk it from 1.56TB to 594GB (-62% size).

Run on a Mac Studio connected with 128GB RAM device. Kimi K3 is the strongest open model to date.

GGUF: unsloth/Kimi-K3-GGUF
Guide: https://unsloth.ai/docs/models/kimi-k3

5 replies

·

danielhanchen

posted an update 11 days ago

Post

4683

Introducing Unsloth for AMD 🚀
You can now train & run LLMs on your AMD hardware

• We collaborated with AMD to enable you to train & run 500+ models on AMD GPUs
• Works on Windows, WSL, Linux
• Train Qwen, Gemma on just 3GB VRAM

GitHub: https://github.com/unslothai/unsloth
Blog + Guide: https://unsloth.ai/docs/basics/amd

3 replies

·

danielhanchen

posted an update 14 days ago

Post

5747

Gemma 4 is now faster and much more accurate! 🚀

Google made huge improvements to tool-calling and chat accuracy, reliability + speed.
To get fixes, re-download our updated GGUF, MLX, NVFP4 quants!

Unsloth quants: https://huggingface.co/collections/unsloth/gemma-4
Gemma 4 Guide: https://unsloth.ai/docs/models/gemma-4

8 replies

·

danielhanchen

posted an update 17 days ago

Post

4656

We’re releasing Gemma 4 NVFP4 quants that run 1.5× faster on your GPU.

Gemma-4-12B NVFP4 works on 11GB VRAM.
26B-A4B hits 13K tok/s (B200).

Unsloth NVFP4 enables faster, more accurate 4-bit Blackwell inference.

Blog: https://unsloth.ai/docs/basics/nvfp4
Gemma NVFP4: https://huggingface.co/collections/unsloth/nvfp4

3 replies

·

danielhanchen

posted an update 22 days ago

Post

4242

We’re releasing new Qwen3.6 quants that run 2.5× faster on your GPU. ⚡

Qwen3.6-27B NVFP4 runs on 24GB VRAM.
35B-A3B can hit 17,561 tok/s (B200).

We also improved accuracy, tool calling, agent use, and looping.

Qwen3.6 NVFP4: https://huggingface.co/collections/unsloth/nvfp4
Guide: https://unsloth.ai/docs/models/qwen3.6#nvfp4

1 reply

·

danielhanchen

posted an update 24 days ago

Post

6155

DeepSeek-V4 can now run locally with Unsloth GGUFs! 🐳

Run lossless DeepSeek-V4-Flash on 168GB RAM or
3-bit works on 110GB Mac, RAM, VRAM setups.

Run via Unsloth Studio or llama.cpp.

GGUF: unsloth/DeepSeek-V4-Flash-GGUF
Guide: https://unsloth.ai/docs/models/deepseek-v4

danielhanchen

posted an update about 1 month ago

Post

3360

1-bit GLM-5.2 GGUF vs. Claude 4.8 Opus vs. GPT-5.5

We gave 3 models the same prompt and compared one-shot outputs.

The 1-bit GLM-5.2 GGUF ran locally on a Mac Studio M3 Ultra with 256GB RAM at ~21.6 tok/s.

Which output do you like best?
GGUF: unsloth/GLM-5.2-GGUF

3 replies

·

danielhanchen

posted an update about 2 months ago

Post

4606

Google's new DiffusionGemma can now run at 2000+ tokens/sec! ⚡

We made local DiffusionGemma inference 1.8× faster.
Run it on 18GB RAM via Unsloth Studio.

GitHub: https://github.com/unslothai/unsloth
Guide: https://unsloth.ai/docs/models/diffusiongemma

4 replies

·

danielhanchen

posted an update about 2 months ago

Post

1193

Google releases DiffusionGemma.✨
The new 26B-A4B diffusion text model runs locally on 18GB RAM.

Run with 4x faster text generation, thinking, image, video and 256K context. Run and train via Unsloth Studio.

GGUF: unsloth/diffusiongemma-26B-A4B-it-GGUF
Guide: https://unsloth.ai/docs/models/diffusiongemma

1 reply

·

danielhanchen

posted an update about 2 months ago

Post

4300

Google releases Gemma 4 QAT. ✨
You can now run Gemma 4 at 3x less memory with near original performance.

QAT makes it possible to run Gemma 4 26B-A4B on 16GB RAM.

GGUFs: https://huggingface.co/collections/unsloth/gemma-4-qat
QAT Guide: https://unsloth.ai/docs/models/gemma-4/qat

1 reply

·

danielhanchen

posted an update about 2 months ago

Post

9351

Gemma 4 12B can now run locally on just 8GB RAM via Dynamic GGUFs.

Google's new model, Gemma 4 12B Unified supports image, audio and 256K context.
You can run and train the model via Unsloth Studio.

GGUF: unsloth/gemma-4-12b-it-GGUF
Guide: https://unsloth.ai/docs/models/gemma-4

5 replies

·

danielhanchen

posted an update 2 months ago

Post

2839

Qwen3.6 MTP is here! Run locally on 20GB RAM. ⚡️

MTP enables Qwen3.6 to generate ~1.4–2.2× faster with no accuracy change.

Qwen3.6-27B: unsloth/Qwen3.6-27B-MTP-GGUF
Qwen3.6-35B-A3B: unsloth/Qwen3.6-35B-A3B-MTP-GGUF
Guide: https://unsloth.ai/docs/models/qwen3.6#mtp-guide

2 replies

·

danielhanchen

posted an update 3 months ago

Post

6001

We’re excited to announce that Unsloth has joined the PyTorch Ecosystem! 🔥🦥

Unsloth is an open-source project that makes training & running models more accurate and faster with less compute. Our mission is to make local AI accessible to everyone. Thanks to all of you for making this possible! 💕

Blog: https://unsloth.ai/blog/pytorch
GitHub: https://github.com/unslothai/unsloth

2 replies

·

qgallouedec

posted an update 3 months ago

Post

10627

Shipped hf-sandbox! 🥡

🧪 Running an eval that executes model-generated C on a few thousand prompts? You probably don't want any of that on your laptop.
Just shipped hf-sandbox, a Modal-style sandbox API on top of Hugging Face Jobs. Spin up an isolated, ephemeral container, run untrusted code, get the result back. No Docker on your laptop, no infra to manage.

Just pip install hf-sandbox.

Early days (v0.1); feedback and issues very welcome:
👉 https://github.com/huggingface/hf-sandbox

1 reply

·

danielhanchen

posted an update 3 months ago

Post

7792

We collaborated with NVIDIA to teach you how we made LLM training ~25% faster! 🚀

Learn how 3 optimizations help your home GPU train models faster:
1. Packed-sequence metadata caching
2. Double-buffered checkpoint reloads
3. Faster MoE routing

Guide: https://unsloth.ai/blog/nvidia-collab
GitHub: https://github.com/unslothai/unsloth

qgallouedec

posted an update 3 months ago

Post

458

**TRL v1.4 is out 🚀** Chunked NLL loss for SFT and a first-class **OpenReward** integration.

**Chunked NLL loss for SFT — drops peak VRAM by up to 14×**

Standard SFT materializes a full [batch × seq × vocab] logits tensor before computing cross-entropy, which dominates peak memory at long context lengths. The new loss_type="chunked_nll" path drops ignored-label tokens before the lm_head matmul and computes cross-entropy in checkpointed chunks of 256.

Peak GPU memory, AdamW fp32:
- Qwen3-14B, 8×H100 FSDP2, 16k seq: 58.9 GB → 38.9 GB
- Qwen3-4B, 1×H100 80GB, 16k seq: OOM → 63.8 GB
- Qwen3-32B, 8×H100 FSDP2, 8k seq: OOM → 71.2 GB

End-to-end it's consistently as fast or faster than nll, and unlocks sequence lengths that don't fit at all under the standard path.

SFTConfig(loss_type="chunked_nll")

Works with PEFT and VLMs out of the box.

**Open Reward Standard environment adapter**

The new trl.experimental.openreward adapter plugs any environment speaking the [Open Reward Standard](https://openrewardstandard.io) protocol into any TRL trainer that takes an environment_factory. One string — a catalog name or a URL — wires the dataset, factory, and reward_func slots; tools are bound dynamically from JSON Schema, no per-env wrapper code:

from trl import GRPOTrainer
from trl.experimental.openreward import OpenRewardSpec

spec = OpenRewardSpec("Eigent/SETA", num_tasks=64)

trainer = GRPOTrainer(
    ...,
    train_dataset=spec.train_dataset,
    environment_factory=spec.environment_factory,
    reward_funcs=spec.reward_funcs,
)

v1.4 also brings MFU helpers for dense + MoE models, GRPO support for Liger 0.8.0 (delta clipping + VESPO + KL bias correction), Tülu 3's length-normalized DPO loss, four more training chat templates (Cohere, Cohere2, Gemma 3, Qwen3-2507), and a 5+ GB CUDA memory leak fix in activation offloading.

Full release notes: https://github.com/huggingface/trl/releases/tag/v1.4.0

danielhanchen

posted an update 3 months ago

Post

8942

We made a guide on how to run open LLMs in Claude Code, Codex and OpenClaw.

Use Gemma 4 and Qwen3.6 GGUFs for local agentic coding on 24GB RAM

Run with self-healing tool calls, code execution, web search via the Unsloth API endpoint and llama.cpp

Guide: https://unsloth.ai/docs/basics/api

danielhanchen

posted an update 3 months ago

Post

10866

Unsloth is now one of the top 10 most followed organizations on Hugging Face. 🤗🦥

Thanks so much for all the support!
Our HF page:

unsloth

5 replies

·

qgallouedec

posted an update 3 months ago

Post

8173

TRL v1.3 ships day-one training support for Qwen 3.6 🚀

The new Qwen 3.6 family (Qwen/Qwen3.6-27B, Qwen/Qwen3.6-35B-A3B) reuses the Qwen3.5-MoE architecture but ships a slightly different chat template, so we updated the stack end-to-end: new training template with {% generation %} markers, tool-call response schema routing, tiny test models for the VLM matrix.

SFT with assistant-only loss works out of the box:

from trl import SFTConfig, SFTTrainer

trainer = SFTTrainer(
    model="Qwen/Qwen3.6-27B",
    args=SFTConfig(assistant_only_loss=True),
    train_dataset=dataset,
)
trainer.train()

So does GRPO tool-calling — just hand tools=[...] to GRPOTrainer.

v1.3 also brings a new experimental TPO trainer (Triple Preference Optimization), speculative decoding in trl vllm-serve (Qwen3 MTP / Eagle3 drafts), 12 more KTO ↔ DPO alignment PRs (KTO promotion to stable is now in reach), three more {% generation %} chat templates (Gemma/Gemma 2, Phi-3, GLM-4-MoE), and a chunky SFT entropy bug fix.

Full release notes: https://github.com/huggingface/trl/releases/tag/v1.3.0

Some Org

AI & ML interests

Team members 55
private

AI & ML interests

Team members 55 private

someorgtoo's activity

Team members 55
private