Xiaomi's affordable flagship AI leads the open models, and Anthropic says Claude helped get it there

AI Digest

小米开源AI模型MiMo-V2.6-Pro以低成本领先性能榜，但被Anthropic指控通过Claude窃取数据训练，凸显开源竞争与伦理争议。

Xiaomi's MiMo-V2.6-Pro leads open AI benchmarks with low cost, but Anthropic accuses it of data theft via Claude, highlighting open-source competition and ethics.

Key points

• MiMo-V2.6-Pro以46分登顶开源模型性能榜，成本仅为竞品1/10 MiMo-V2.6-Pro scores 46 on Intelligence Index, outperforming rivals like Qwen
• 强化学习三轴扩展使模型性能提升显著，但被指窃取Claude数据 Reinforcement learning scaling boosted performance but faced data theft accusations
• 公司开放RL工具链与7000个自动评分训练任务 Xiaomi opens RL toolkit with 7,000 auto-graded training tasks
• Anthropic指控小米等7家机构非法蒸馏Claude数据 Anthropic names Xiaomi among 7 labs accused of illegal distillation
• 模型成本优势使其在性价比上占据Pareto前沿 Cost efficiency places it at the Pareto frontier of AI value

Takeaway: 开源AI竞争加剧，但数据伦理争议可能影响技术发展路径。 / Open-source AI competition intensifies, but data ethics disputes could reshape technological trajectories.

Why it matters 揭示AI行业开源竞争与数据伦理的双重挑战，具现实警示意义。