📺 Xiaomi Is Livestreaming a Production RL Run — Burn Rate Included @XiaomiMiMo is training its MiMo-V2.6 in public — not a demo, a live post-training dashboard where every step's reward, loss and benchmark movement is watchable in real time. One day in, the meter reads roughly $1.2M spent and 60B tokens consumed. As Zhihu contributor Kitt在进化 puts it: this isn't a training site, it's a money-burning site — about ¥4,000 a minute. His bigger point: this is rare, valuable data. Almost nobody publishes what frontier-scale RL actually costs. 1️⃣ What the dashboard actually shows Every training step exposes the internals most labs keep private. His reading guide: 🔹 Loss family: training loss, entropy (are answers diversifying or collapsing into one mode), gradient norm, and train/inference KL divergence. 🔹 Reward: a mix of test-case scoring and rubric-based reward. 🔹 dynsam (dynamic sampling): avg@n performance over repeated tries, plus pass-rate buckets — including samples never solved and always solved. 🔹 DeepSWE v1.1 as the headline probe (mini-swe-agent, avg@3): Pro at 63.7, Flash at 60.7 at the time of writing. 2️⃣ The cost ledger, out in the open From the dashboard he extracts the numbers the industry usually guesses at: 🔹 MiMo-V2.6 Pro: roughly $36 per million tokens of training. 🔹 MiMo-V2.6 Flash: roughly $8 per million tokens. 🔹 His rule of thumb: equivalent inference runs 30-50x cheaper than these training figures. 3️⃣ What's actually being trained The sample distribution is also public: about 1,500 prompts per step, ~70% of them coding tasks. The remaining third is split across general, visual, cybersecurity and chat tasks — which makes this, in effect, an agentic-coding-centric RL run with side dishes. 4️⃣ Why it's worth watching He notes a university lab livestreamed a training run days earlier, but at nowhere near this scale or frontier relevance. For anyone who wants to learn how production RL behaves — rewards, entropy, dynamic sampling, benchmarks moving step by step — this is a rare open classroom. His only complaint, half-joking: a livestream this good deserves a comment section. 🔗 Full Reading: #Xiaomi# #MiMo# #ReinforcementLearning# #PostTraining# #LLM# #OpenScience# #AI#

Just returned from an intense but enlightening week in China, visiting the major AI labs, robotics startups and EV manufacturers. Spent time with old friends and new ones including @lexfridman, @natolambert, @xeophon and the @readsail team. Blessed to have met with the leaders, researchers and engineers defining the future of open source and AI and robotics including @kaifulee, @yaqinzhang, @Wang XingXing (@UnitreeRobotics CEO) and researchers and founders from @deepseek_ai, @Zai_org, @Kimi_Moonshot, @01.ai, @alibaba_cloud @Alibaba_Qwen, @MiniMax_AI, @BytedanceTalk, @XiaomiMiMo, @xiaohongshu, @GalbotRobotics, @UnitreeRobotics, @Galaxea_x, @AntGroup (@AntLingAGI) and others. Stay tuned for my blog post on insights, takeaways and company spotlights, coming soon!

Xiaomi MiMo-V2.6 is now open—a native multimodal agent family built for large-scale reinforcement learning. 🚀📜 MIT License. 🤖 🏆 MiMo-V2.6-Pro scores 46 on the Artificial Analysis Intelligence Index and reaches 71.9 on DeepSWE v1.1, 89.9 on Terminal-Bench 2.1, and 82.0 on OSWorld-Verified. 🧠 The 1.02T MoE activates 42B parameters and supports text, images, video, audio, and a 1M-token context. ⚙️ One mixed RL run trains coding, general, visual, and cybersecurity agents together. Pro and Flash completed 30 steps each in under six days, producing around 750K trajectories. 🌐 The models support computer use, 3D creation, embodied control, coding, design, video, and music workflows.

Xiaomi MiMo-V2.5-Pro achieves multiple breakthroughs in the latest Arena rankings (Apr 26, 2026) 🔥 🏆 Text Arena (Expert) — #6# globally | #1# open-source model Also #1# among Chinese models, with Xiaomi ranking #3# globally by lab, behind only Anthropic and OpenAI. Expert is defined by high-difficulty tasks and expert voting, measuring core model intelligence. 🏆 Text Arena (Overall) — #2# open-source globally Strong across math, coding, creative writing, and general text tasks. 🏆 Code Arena (WebDev) — #3# open-source globally Evaluated by real community blind voting on frontend code generation. 🏆 Text Arena sub-rankings — #1# open-source globally in 4 categories Hard Prompts, Hard Prompts(English), Instruction Following and Long Query. Real-world preference, real model strength.

📢Calling all Apache Software Foundation committers Xiaomi MiMo is giving you our Max Token Plan for FREE as part of the 100T Token Grant for Builders Program. Sign up with your email → instantly activated. Sign up now:

Today we're excited to announce the first 13 ecosystem partners joining the Xiaomi MiMo Orbit Program. A sincere thank you to every partner for your trust and collaboration. @gitlawb and 12 other ecosystem partners. This is just the beginning. MiMo Orbit Program is still open, and we'd love to collaborate with you! → business-mimo@xiaomi.com

Heads up, agent users！ If you're using Xiaomi MiMo with thinking mode: When thinking mode is enabled in a multi-turn agent session and the conversation history contains a tool call, any assistant message with tool calls passed back in subsequent user turns must preserve its full reasoning_content field — otherwise the API will return a 400 error. Without it, the model's context is incomplete, which can lead to weaker instruction-following, more hallucinations, and a visibly degraded user experience. Missing reasoning = incomplete context = degraded reasoning quality. Affected frameworks include TRAE, Cursor, Roo Code, Codex, GitHub Copilot CLI, Zed, AutoGen. We're actively working with the maintainers to push compatibility updates. Affected models: MiMo-V2.5-Pro, MiMo-V2.5, MiMo-V2-Pro, MiMo-V2-Omni, MiMo-V2-Flash. See docs( )for more details.

Grok this, Opus that. GPT resets. Whatever man. Ran Xiaomi Mimo-v2.6 yesterday with @drost_ai. 2-hour engagement. Root access through an authenticated endpoint in a container, lateral movement, root access on another container, database access and dumped the database. 50 cents.