Image created with gemini-2.5-flash-image with claude-sonnet-4-5. Image prompt: Seamless repeating wrapping paper pattern featuring vintage cartographic elements including stylized map borders, compass roses, latitude lines, diplomatic wax seals in burgundy and gold, international passport stamps, treaty ribbons, fragments of multilingual text, and the word INTERNATIONAL as elegant banner typography woven throughout, Victorian decorative arts style, deep navy and antique gold with cream accents, sophisticated gift wrap quality with embossed texture feel, timeless diplomatic elegance
MiniMax M2.1: Significantly Enhanced Multi-Language Programming, Built for Real-World Complex Tasks – MiniMax News https://www.minimax.io/news/minimax-m21
Shoutout @theo for today’s live stream using M2.1! 🙌 Loved watching a real dev put it through in real-time! Key Takeaways: “M2.1 is really good at long-horizon tasks and could generate surprisingly good results” Best part? It’s 1/10 the price of Opus. Ready to be surprised? https://x.com/MiniMax_AI/status/2003673337671602378
A seven to eight month gap between released closed US models and released Chinese open weights models continues to be a good rule of thumb, as it has been all year.”” / X https://x.com/emollick/status/2003217274510143709
China issues draft rules to regulate AI with human-like interaction | Reuters https://www.reuters.com/world/asia-pacific/china-issues-drafts-rules-regulate-ai-with-human-like-interaction-2025-12-27/
Structured for intelligence: Why Al needs governed, discoverable, and provisioned data | dbt Labs https://www.getdbt.com/resources/structured-for-intelligence-why-al-needs-governed-discoverable-and-provisioned-data
@vllm_project @HotAisle I got MiniMax-M2.1 running on my mi300x with sglang. (I had to patch it) FP8: 55 TPS bf16: 71 TPS Still showing the same behavior where FP8 perf is worse, rather than better. so this is across both vLLM and sglang, the implementation of FP8 on mi300x is slower than bf16.”” / X https://x.com/QuixiAI/status/2005724765928210655
@vllm_project @HotAisle MiniMax-M2.1 on VLLM (FP8) gets 42 TPS. I’ll next try converting it to bf16 and see how fast it is”” / X https://x.com/QuixiAI/status/2005481942712811695
@vllm_project @HotAisle MiniMax-M2.1-bf16 gets 55.7 TPS. https://x.com/QuixiAI/status/2005502089653547174
@vllm_project @HotAisle The patch is here: https://x.com/QuixiAI/status/2005746928399827407
🚨Code Arena Update Minimax-M2.1 debuts at #1 open model on WebDev leaderboard and lands #6 overall with a 1445 score, tying with GLM-4.7. These scores come from Code Arena, where models build websites, apps, and games from a single prompt. Congrats @MiniMax__AI on this latest https://x.com/arena/status/2005779347182084585
Excited to have builders to experience M2.1 on Kilo! Let’s code smarter.”” / X https://x.com/MiniMax_AI/status/2003606223191703708
Have you checked out the new open model MiniMax M2.1 by @MiniMax__AI in the Code Arena? 👀 The next generation of live coding evals happens in the Code Arena. Built to test how models plan, scaffold, debug, and build real web apps step-by-step. Remember, your votes drive the”” / X https://x.com/arena/status/2003585316029104383
MiniMax M2.1 is live in Kilo – by Ari – Kilo Blog https://blog.kilo.ai/p/minimax-m21
MiniMax M2.1 is live on Chutes 🪂 We also ran the official Provider Verifier, results: • 100% query success • 82.83% tool calling • 95.12% tool accuracy (4 edge cases) • 100% response quality • 100% language following. Try now: https://x.com/chutes_ai/status/2005539785923072424
Seeing @dhh share his experience was both humbling and deeply encouraging! Knowing that MiniMax M2.1 can meaningfully contribute on codebases as large and complex as Ruby on Rails is a strong signal we’re moving in the right direction. It’s still far from perfect and we have to”” / X https://x.com/MiniMax_AI/status/2005536770226811014
Super excited to see Minimax 2.1 posting good results on our SWE-bench Verified, SWE-bench Multilingual and SciCode benchmarks! https://x.com/OfirPress/status/2003625671042732329
This one’s for builders, not demos. MiniMax M2.1 is now live for 30M devs on @blackboxai — appreciate the team 🙌 Let’s go further!”” / X https://x.com/MiniMax_AI/status/2003926396335460447
Want to use MiniMax M2.1 in Roo Code? It’s here for you!”” / X https://x.com/MiniMax_AI/status/2003611728320561528
We’ve evaluated @MiniMax__AI’s M2.1 on our Vals Index. Among open-weight models, it ranks #2 behind GLM 4.7, but with a lower latency and cost compared to GLM 4.7. https://x.com/ValsAI/status/2003646964664287667
A significant stride in India’s space sector… The successful LVM3-M6 launch, placing the heaviest satellite ever launched from Indian soil, the spacecraft of USA, BlueBird Block-2, into its intended orbit, marks a proud milestone in India’s space journey. It strengthens https://x.com/narendramodi/status/2003677923820335183
Powered by India’s youth, our space programme is getting more advanced and impactful. With LVM3 demonstrating reliable heavy-lift performance, we are strengthening the foundations for future missions such as Gaganyaan, expanding commercial launch services and deepening global https://x.com/narendramodi/status/2003681952323502169
2025 exposed 3 uncomfortable truths about open-source AI: 1. Leadership requires sustained commitment E.g., Meta’s wavering on Llama 4 and Behemoth created an opening that Chinese rivals filled. You can’t dominate open source part-time or as a hedge 2. “Open” now exists on a https://x.com/TheTuringPost/status/2001769019096056302
strong new open 32B VLM model from Korea, which has good english and very korean benchmark scores! the Artificial Analysis score is in part due to a very high taubench score, but other bench are good as well + vision understanding also seems strong! now the fun part while https://x.com/eliebakouch/status/2005549508063559876
Leave a Reply