MiMo-V2.6-Pro-RL: Xiaomi multimodal reasoning model
Xiaomi's new RL-trained flagship: vision-language, audio and video understanding with long context and agent tuning.
Why it matters
MIT-licensed open weights; hit HuggingFace's top-15 trending within days of release (Sep 21).
What you could build with it
Benchmark it as an open agent backbone for multi-step research tasks against Claude/GPT.
Does it hold up?
Close to the frontier on agent benchmarks at a fraction of the token price, but the builder verdict is that self-hosting the -RL checkpoints needs a serious GPU node — most builders will realistically reach it via Xiaomi's API or OpenRouter; independent hands-on testing is only days old.
Built with MiMo-V2.6-Pro-RL
- Kingy.ai: MiMo-V2.6-Pro Benchmarks & Hands-On Test vs Claude Opus 5kingy · Independent benchmark roundup plus Kingy.ai's own hands-on test of Pro vs Flash vs Claude Opus 5 on three OpenCode agent tasks: top open-weights model at 46 on the AA index, but 5-7 points behind the closed frontier, and the per-token price edge shrinks because the model is verbose.
- OrcaRouter: Xiaomi MiMo-V2.6-Flash — a 309B Open Model to Serveorcarouter · Deployment-focused third-party analysis of the -RL checkpoints: documentation drift between model card and config, FP8 weight sizing, serving recipes (SGLang tp16/dp2, vLLM tp8), and the honest framing that open weights removed the license barrier, not the hardware barrier.
- AI Frontier Post: Xiaomi open-sources MiMo-V2.6dev.to · Third-party writeup on the live-streamed RL run, the 7,000+ task environments and RL code, unchanged V2.5 pricing, and the caveat that the 46.32 Intelligence Index score is vendor-reported.
First spotted on huggingface: source.
More AI models
ElevenLabs Eleven v4 + v4 Turbo: new emotive TTS architecture with a 100ms real-time variantElevenLabs released Eleven v4 and Eleven v4 Turbo on September 28: a new text-to-speech architecture with inline…models · JEV 0.72VoiceStudio: open-source, fully-local ElevenLabs alternative with voice cloning, dubbing and dictation in 646 languagesVoiceStudio is an open-source, fully-local voice AI studio: voice cloning from a 3-second sample, voice design, video…models · JEV 0.71DeepSeek V4-Flash official API: public beta with upgraded agent capabilities, Responses API and Codex supportDeepSeek's official V4-Flash is now a public beta API with massively upgraded agent capabilities - benchmark scores…models · JEV 0.69Mistral Forge: enterprise platform for training and continuously improving proprietary modelsMistral AI launched Forge (Sep 28) — a full-lifecycle model training platform (pre-training, SFT, DPO/ODPO, RL…models · JEV 0.69kev: open, trainable Jev-like family of small decision models on Qwen3.5/3.8A family of small decision models (0.8B to 27B) built on Qwen3.5/3.8 that reproduces Jev's prefill-only typed-decision…models · JEV 0.68GPT-Synopsys: OpenAI and Synopsys multi-year deal to build an AI model that operates EDA toolsOpenAI and Synopsys signed a multi-year agreement on Sep 30 to jointly develop GPT-Synopsys, a specialized model that…models · JEV 0.66
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.