🤯 Xiaomi showed a 150W mini AI box designed to run a 120B model LOCALLY. Yeah, it's Xiaomi, so good luck getting it in the US if it ever ships. The Xiaomi AI Cube, a prototype built around 3 of Xiaomi's own XRING chips. And the specs are kind of crazy 👉 🧠 Local deployment → 120B + 3B models 💾 D100 → up to 160GB unified memory 👀 🚀 O100 → 1.22 TB/s near-memory bandwidth 👀 Nvidia did you see this? ⚡ Entire AI Cube → up to 150W 🧮 O3 → 200 TOPS NPU 🎮 O3 → 16-core G2 Ultra NX GPU That 1.22 TB/s number is especially interesting. According to Xiaomi, they stack high-speed DRAM directly over the O100's logic/NPU layer using wafer-on-wafer packaging + hybrid bonding. In other words 👉 very short path between memory and compute. And Xiaomi says the D100 itself can accommodate local models as large as 200B parameters. 👀 🎯 Strix Halo showed what 128GB unified memory could do. Gorgon Halo is up next with 192GB. Now Xiaomi is experimenting with 160GB-class unified memory + >1 TB/s bandwidth in a 150W mini box. Caveats of course ... ⚠️ It's a prototype ⚠️ No price ⚠️ No retail release date ⚠️ O100/D100 commercial use is planned for 2027 🔗 Source: GizmoChina

📺 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#

BillionLive Creator Incentive Program: Week 3 Winners Announced 🎬 The BillionLive Creator Weekly Leaderboard Week 3 has concluded! Big thanks to our creators for the deep insights and the on-chain grind. Leaderboard: 🥇 Molly @Molly9975019573 (238 USDT + 3,000 Billion Points) 🥈 Cc Egg-Fried-Rice @Cc910813 (138 USDT + 2,000 Billion Points) 🥉 Shuishui (100 USDT + 1,200 Billion Points) 4️⃣ Franky @franky7 (60 USDT + 900 Billion Points) 5️⃣ Ashu (30 USDT + 900 Billion Points) 6⃣-🔟 Moe, Mubai, Duoduo, Xinxin, and Xiaomiaomiao (Each receives 20 USDT + 550 Billion Points) Highlights: Molly dived deep into @fomo mechanics, while Cc and Shuishui mastered the patient art of trenches. Week 4 is now live. The weekly Top 10 will continue to share 666 USDT + 10,750 Billion Points. Event Page: BillionLive 创作者激励计划第三期优胜者公布🎬 BillionLive 创作者周榜第三期圆满结束！感谢本期创作者们带来的精彩直播与深度分享。 本期优胜榜单： 🥇 Molly @Molly9975019573 (238 USDT + 3,000 Billion Points) 🥈 Cc 蛋炒饭 @Cc910813 (138 USDT + 2,000 Billion Points) 🥉 水水 (100 USDT + 1,200 Billion Points) 4️⃣ Franky @franky7 (60 USDT + 900 Billion Points) 5️⃣ 阿树 (30 USDT + 900 Billion Points) 6⃣-🔟 打三角洲的一只莫、沐白、多多、心心、小淼淼 (每人获得 20 USDT + 550 Billion Points) 本期切片高光： Molly 深度拆解 @fomo 玩法策略，Cc 蛋炒饭 和水水聊链上枯坐打狗，实盘操作。 第四期创作者周榜现已开启。 每周 Top 10 将继续瓜分 666 USDT + 10,750 Billion Points。 活动链接：

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!