---
base_model:
- Qwen/Qwen3.5-0.8B
language:
- fr
library_name: transformers
pipeline_tag: text-generation
tags:
- french
- conversational
- edge
- qwen
license: apache-2.0
datasets:
- kurakurai/Luth-2-Post-Training-SFT
- kurakurai/Luth-2-Post-Training-RL
---


![luth-2-banner](https://cdn-uploads.huggingface.co/production/uploads/66beff523ae330ae8b698ee9/wso_h0n6syZf9E8YDKWYA.png)

# Luth-2-0.8B

**Luth-2-0.8B** is a 750M-parameter (text only) non-reasoning model, setting a new **state of the art in French** for its size across math, code, instruction following, general knowledge and tool calling. It is trained on a 3B-token French SFT mixture followed by multi-domain on-policy distillation (MOPD). The model outperforms every other model in its size class on our selected French benchmarks and stays competitive with models 2 to 3 times larger. It is small enough for efficient local and on-device deployment.

- 📄 **Blog**: [Luth-2: Pushing the French Capabilities of SLMs with MOPD](https://huggingface.co/blog/MaxLSB/luth-2)
- 🤗 **Models**: [Luth-2-0.8B](https://huggingface.co/kurakurai/Luth-2-0.8B) · [Luth-2-2B](https://huggingface.co/kurakurai/Luth-2-2B)
- 📊 **Datasets**: [SFT](https://huggingface.co/datasets/kurakurai/Luth-2-Post-Training-SFT) · [RL](https://huggingface.co/datasets/kurakurai/Luth-2-Post-Training-RL)
- 💻 **Code**: [GitHub](https://github.com/kurakurai/Luth-2)
- 🏆 **Leaderboard**: [French LLM Leaderboard](https://huggingface.co/spaces/kurakurai/llm_leaderboard_fr)  

![luth2_benchmarks_0.8b_portrait](https://cdn-uploads.huggingface.co/production/uploads/66beff523ae330ae8b698ee9/gER5neg6becY6FxAeXFSG.png)

> [!NOTE]
> **Luth-2-0.8B** inherits the VLM architecture of Qwen3.5-0.8B but was not trained on vision data. We do not recommend using it for vision tasks.

## Model variants

| Model | Description |
|---|---|
| [Luth-2-0.8B](https://huggingface.co/kurakurai/Luth-2-0.8B) | Original checkpoint in native format. Best for fine-tuning or inference with Transformers, vLLM and SGLang. |
| [Luth-2-0.8B-GGUF](https://huggingface.co/kurakurai/Luth-2-0.8B-GGUF) | Quantized format for llama.cpp and compatible tools. Optimized for CPU inference and reduced memory usage. |

## Training

**Luth-2-0.8B** is post-trained from Qwen3.5-0.8B in two stages:

1. **Supervised fine-tuning** on [Luth-2-Post-Training-SFT](https://huggingface.co/datasets/kurakurai/Luth-2-Post-Training-SFT), a 3B-token French mixture spanning math (37.2%), knowledge (27.9%), code (22.2%), instruction following (6.5%) and tool calling (6.3%). Prompts were translated from English SFT datasets and answers regenerated with strong open-source teachers.
2. **Multi-domain on-policy distillation (MOPD)**. Three specialists (math, code, instruction following) are trained separately with GRPO on [Luth-2-Post-Training-RL](https://huggingface.co/datasets/kurakurai/Luth-2-Post-Training-RL), then distilled back into the SFT student.

## Inference

**Luth-2-0.8B** is supported by Transformers, vLLM, SGLang and more.

Quick start with Transformers:

```python
from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer

model_id = "kurakurai/Luth-2-0.8B"
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="auto",
    dtype="bfloat16",
    # attn_implementation="flash_attention_2"  # uncomment on compatible GPU
)
tokenizer = AutoTokenizer.from_pretrained(model_id)
streamer = TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)

prompt = "Quelle est la capitale de la France?"
input_ids = tokenizer.apply_chat_template(
    [{"role": "user", "content": prompt}],
    add_generation_prompt=True,
    return_tensors="pt",
    tokenize=True,
)["input_ids"].to(model.device)

output = model.generate(
    input_ids,
    do_sample=True,
    temperature=0.8,
    top_p=0.95,
    top_k=20,
    max_new_tokens=512,
    streamer=streamer,
)
```

## Evaluation

Evaluations can be reproduced using our [GitHub](https://github.com/kurakurai/Luth-2) repository. The benchmarks are French subsets or verified translations, scored with `temperature=0.6, top_p=0.95, top_k=20`, thinking disabled, averaged over 10 runs.

| French Benchmarks | Luth-2-0.8B | Luth-0.6B-Instruct | Qwen3.5-0.8B |
|---|--:|--:|--:|
| MGSM-rev2 | **72.92** | 58.52 | 35.20 |
| AIME 24 | **5.67** | 2.00 | 1.00 |
| AIME 25 | **8.67** | 1.33 | 0.33 |
| Math-500 | **57.60** | 44.74 | 27.46 |
| Global-MMLU-Lite | **53.30** | 40.20 | 44.00 |
| MMLU-ProX-Lite | **38.93** | 25.40 | 27.60 |
| GPQA-Diamond | **26.87** | 25.60 | 23.80 |
| IFEval | **71.23** | 51.23 | 44.47 |
| Multi-IF | **61.52** | 33.77 | 32.72 |
| HumanEval+ | **46.81** | 30.25 | 10.87 |
| MBPP+ | **42.33** | 34.74 | 18.20 |
| BFCL v2 | **64.02** | 61.72 | 51.49 |

See the [French LLM Leaderboard](https://huggingface.co/spaces/kurakurai/llm_leaderboard_fr) for comparisons across models.

## Contact

Questions or feedback? Reach us on LinkedIn: [Maxence Lasbordes](https://www.linkedin.com/in/maxence-lasbordes/) and [Guillaume Pradel](https://www.linkedin.com/in/guillaume-pradel/).

## Citation

```bibtex
@misc{luth2,
  title  = {Luth-2: Pushing the French Capabilities of SLMs with MOPD},
  author = {Maxence Lasbordes and Guillaume Pradel},
  year   = {2026},
  url    = {https://huggingface.co/blog/MaxLSB/luth-2}
}
```