XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B · Hugging Face

Source: https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B MiMo-V2.6-Distill-Qwen-9B is a 9B agentic model developed by Xiaomi MiMo through supervised fine-tuning ofQwen3.5-9Bon MiMo-generated data. It covers coding, general-purpose agent tasks, visual coding, and cybersecurity. We release this SFT checkpoint as a starting point for open research in agentic reinforcement learning.

https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B#evaluationEvaluation

Results for the released SFT checkpoint, as reported in the MiMo-V2.6 technical report.

DomainBenchmarkMetricQwen3.5-9BMiMo-V2.6-Distill-Qwen-9B (SFT)CodeSWE [email protected]61.1CodeSWE [email protected]44.6CodeMiMo Code (mini)†[email protected]51.6CyberMiMo Cyber (mini)†[email protected]31.3GeneralAutomationBench [email protected]30.3GeneralTerminal Bench [email protected]37.1[email protected]35.2[email protected]19.5[email protected]18.3GeneralMiMo General (mini)†[email protected]62.2VisualMiMo Visual Coding (mini)†[email protected]64.0 † Internal evaluation sets.

https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B#training-dataTraining Data

The weighted SFT data mixture contains 77.4B total tokens, including 27.2B loss-bearing tokens.

DomainTotal tokens (B)Token share (%)Loss-bearing tokens (B)Code23.229.97.3Cyber11.014.24.8General22.028.55.7Visual21.227.49.4Total77.4100.0****27.2

https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B#quickstartQuickstart

For text generation, use a recentSGLangbuild with Qwen3.5 support. The checkpoint includes its tokenizer and MiMo v2.6 chat template.

sglang serve \
  --model-path XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B \
  --reasoning-parser mimo \
  --host 0.0.0.0 \
  --port 30000

Query the endpoint with thinking explicitly enabled:

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:30000/v1",
    api_key="EMPTY",
)

response = client.chat.completions.create(
    model="XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B",
    messages=[
        {"role": "user", "content": "What is 15% of 240?"}
    ],
    max_tokens=2048,
    extra_body={"chat_template_kwargs": {"enable_thinking": True}},
)

message = response.choices[0].message
print("Thinking:", getattr(message, "reasoning_content", "") or "")
print("Answer:", message.content or "")

https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B#citationCitation

@misc{mimo2026v26,
  title={MiMo-V2.6: Scaling Reinforcement Learning Towards Self-Improvement},
  author={{Xiaomi MiMo Team}},
  year={2026},
  howpublished={\url{https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Pro-RL}},
}