CoolFace
Modelpublic

kurakurai/Luth-2-0.8B

sourceHugging Faceapache-2.0updated 1mo agoView on Hugging Face
6likes482downloads
Model Card

luth-2-banner

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.

luth2_benchmarks_0.8b_portrait

[!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

ModelDescription
Luth-2-0.8BOriginal checkpoint in native format. Best for fine-tuning or inference with Transformers, vLLM and SGLang.
Luth-2-0.8B-GGUFQuantized 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. 1.Supervised fine-tuning on 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. 2.Multi-domain on-policy distillation (MOPD). Three specialists (math, code, instruction following) are trained separately with GRPO on 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 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 BenchmarksLuth-2-0.8BLuth-0.6B-InstructQwen3.5-0.8B
MGSM-rev272.9258.5235.20
AIME 245.672.001.00
AIME 258.671.330.33
Math-50057.6044.7427.46
Global-MMLU-Lite53.3040.2044.00
MMLU-ProX-Lite38.9325.4027.60
GPQA-Diamond26.8725.6023.80
IFEval71.2351.2344.47
Multi-IF61.5233.7732.72
HumanEval+46.8130.2510.87
MBPP+42.3334.7418.20
BFCL v264.0261.7251.49

See the French LLM Leaderboard for comparisons across models.

Contact

Questions or feedback? Reach us on LinkedIn: Maxence Lasbordes and 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}
}