关注 AI 研究者、开发者与机构的动态
Introducing MiniMax M3: The First Open-Weights Model to Combine Three Frontier Capabilities - Coding & Agentic Frontier: 59.0% SWE-Bench Pro, 66.0% Terminal Bench 2.1, 34.8% SWE-fficiency, 28.8% KernelBench Hard, 74.2% MCP Atlas - MiniMax Sparse Attention scales context to 1M - Natively Multimodal from Step Zero API: platform.minimax.io Token Plan: platform.minimax.io/subscrib… 🚀New! MiniMax Code: code.minimax.io Weights & Tech Report in ~10 Days
客户端：github.com/op7418/CodePilot/…
跟祖传的 20K context 说 bye bye 了。 MiniMax M3 发布了，三个亮点： 1M context、原生多模态、Agentic。 我这次做了一次完整评测，使用CC workflow 、 @ZenMuxAI和MiniMax M3： 给一张截图，做一个“凡人修仙剑阵对决手势游戏”。 要求是：支持双人对决 、使用 workflow 拆解任务、加入石头剪刀布机制。 2 小时后，游戏真的跑起来了。 这一代LLM的版本答案我知道了： 1M 上下文 + 多模态+ agent 模式。 1M context 是推理深度的基础，多 agent 负责拆任务和执行。 Video
推荐理由：官方给出 M3 的开源权重与大上下文等信息，引用评测展示了多模态加 Agent 工作流的实际产出。
MiniMax M3 @MiniMax_AI is now live on Happycapy 🎉 A major upgrade for agent workflows, especially when the task is messy, multimodal, and too large for a normal chat. Now you can give the agent more context upfront and let it work across everything in one run: • Native multimodal input: PDFs, videos, images, screenshots, and long docs • Strong coding and agent work: useful for repo level debugging, issue tracing, and Connect My Mac tasks • Open weights and about 3x cheaper than Sonnet. Build more, pay less! We put it to the test below 🧵 Video
MiniMax M3 已上线 Happycapy，可处理 PDF、视频、图片、截图和长文档等多模态输入，面向体量大且繁琐的智能体工作流。
Introducing MiniMax M3: The First Open-Weights Model to Combine Three Frontier Capabilities - Coding & Agentic Frontier: 59.0% SWE-Bench Pro, 66.0% Terminal Bench 2.1, 34.8% SWE-fficiency, 28.8% KernelBench Hard, 74.2% MCP Atlas - MiniMax Sparse Attention scales context to 1M - Natively Multimodal from Step Zero API: platform.minimax.io Token Plan: platform.minimax.io/subscrib… 🚀New! MiniMax Code: code.minimax.io Weights & Tech Report in ~10 Days
Introducing MiniMax M3: The First Open-Weights Model to Combine Three Frontier Capabilities - Coding & Agentic Frontier: 59.0% SWE-Bench Pro, 66.0% Terminal Bench 2.1, 34.8% SWE-fficiency, 28.8% KernelBench Hard, 74.2% MCP Atlas - MiniMax Sparse Attention scales context to 1M - Natively Multimodal from Step Zero API: platform.minimax.io Token Plan: platform.minimax.io/subscrib… 🚀New! MiniMax Code: code.minimax.io Weights & Tech Report in ~10 Days
推荐理由：flowith 上线 MiniMax M3 支持，读者可了解该模型在 1M 上下文与原生多模态上的接入进展。
Introducing MiniMax M3: The First Open-Weights Model to Combine Three Frontier Capabilities - Coding & Agentic Frontier: 59.0% SWE-Bench Pro, 66.0% Terminal Bench 2.1, 34.8% SWE-fficiency, 28.8% KernelBench Hard, 74.2% MCP Atlas - MiniMax Sparse Attention scales context to 1M - Natively Multimodal from Step Zero API: platform.minimax.io Token Plan: platform.minimax.io/subscrib… 🚀New! MiniMax Code: code.minimax.io Weights & Tech Report in ~10 Days
Chinese AI labs are not slowing down MiniMax M3 just dropped and it’s looking good: > 1M context > native multimodal > super cheap with 50% launch discount > strong coding benchmarks > beats GPT-5.5 and Gemini 3.1 Pro on SWE-Bench Pro gonna test it this week and share my experience which model should i compare it with first?
英伟达发布基于 ARM 的 RTX Spark，配备 RTX 5070 级 GPU、128GB 统一内存和 1 PetaFLOP 的 FP4 算力，可在本地运行 200B 参数大模型。
NVIDIA, ARM tabanlı yeni işlemcisi RTX Spark'ı duyurdu. - İşlemcide RTX 5070'e denk bir GPU bulunuyor. - Modern oyunlarda 1440P'de 100 FPS'te çalışıyor. - Laptop, Windows olmasına rağmen prizden çektiğinizde performans düşmüyor. - Batarya ömrü uzun. - Sadece laptoplar için değil masaüstü bilgisayarlarını da hedefliyor. - Sahnede 007 First Light ve Forza Horizon 6 ile gösterildi. - Yapay zeka işlem gücü de yüksek. - 2026 Sonbahar'ında çıkacak. Video
MiniMax M3 might be the most underrated coding model right now. I gave it nothing but a screenshot of a chaotic 90s GeoCities-style fan page, no HTML source, just the image + the asset files, and told it to rebuild the whole thing as a sleek Apple-style 2026 site. One shot. Through OpenCode. The result is genuinely stunning. It kept the soul (the "stevibe's HyperHome" identity, the visitor counter, the guestbook, the webmaster portrait) and translated every section into clean modern design, gradient hero, proper typography, dark theme, the works. Video
Minimax M3 is excellent at SVG generation, reaching close to Gemini 3.5 Flash levels and beating Opus 4.7 on SVG-Bench. With 1M context, native multimodality, strong agentic/coding ability and open weights coming soon, the closed-source moat is thinning fast. Full Video: piped.video/p6Npi-HBoRU Video
免费领6个月ChatGPT Pro， 价值$ 1200🤩 这可能是今年对开发者最实在的福利了， 没有硬性Star数要求， 有项目链接基本都能过， 只要你是任何一个公开开源项目的维护者，哪怕只有你一个人在维护， 都可以去申请试试： openai.com/form/codex-for-os…
🚀 We’re launching MiniMax M3 from @MiniMax_AI on Novita AI as a Day-0 API launch partner. The first open-weights model to combine: • Frontier coding and agent capabilities 59.0% SWE-Bench Pro, 66.0% Terminal Bench 2.1, 74.2% MCP Atlas • Up to 1M tokens of context Enabled by MiniMax Sparse Attention • Native multimodality from day one Built for both text and visual understanding Excited to make MiniMax M3 available to developers on Day 0. Video
推荐理由：M3 在 Novita 上线并附上 SWE-Bench Pro 等基准，可据此了解其编码与智能体能力定位。
MiniMax M3 scores 90.3% GPT 5.5 Scores 92.4% Just a 2.1% gap now at @convex. Incredible to see the open-source models closing in. The future of AI is open. 🤖💪
KwaiKeye 在 Hugging Face 发布多模态模型 Keye VL 2.0-30B-A3B，总参数 30B、活跃参数 3B，采用 Apache 2.0 完全开源。
Keye VL 2.0-30B-A3B 🔥 New multimodal model from @KwaiKeye ✨ 30B/3B active - Apache 2.0 ✨ 256K context via DeepSeek Sparse Attention (probably the first model to ship this in production?👀) ✨ Gets MORE accurate as you feed it more frames ✨ Matchs Qwen3 VL and Gemini 3 Flash on benchmarks
免费领6个月ChatGPT Pro， 价值$ 1200🤩 这可能是今年对开发者最实在的福利了， 没有硬性Star数要求， 有项目链接基本都能过， 只要你是任何一个公开开源项目的维护者，哪怕只有你一个人在维护， 都可以去申请试试： openai.com/form/codex-for-os…
很高兴与 @BAI_AGI 联手 🤝 M3 提供大脑，现在智能体有了可以运作的资金通道🔥
B.AI 🤝 @MiniMax_AI Powerful agents require both advanced intelligence and robust economic infrastructure. We are partnering with MiniMax to help accelerate the future of AI-native applications and autonomous agents. As MiniMax advances the frontier of reasoning, coding, multimodal understanding, and agentic workflows, B.AI is building the financial infrastructure that enables intelligent agents to transact, coordinate, and operate within decentralized economies.
小小任务拿捏😄，前提是你需要配置一些可以抓东西的工具就行了。 比如搜索、抓取 这类的之前已经推荐过好几次了，可以看看。
这种小任务爬爬虾的工作交给Qclaw这种还是挺好的，每日有4000K的Tokens可以用啊。 不用白不用啊，整理成报告还是可以滴。不然你不用也浪费了啊！
🧵 1/8 前几天给大家分享的我用AI邮箱来搞钱的帖子(见前贴)，有很多朋友私信我问，你使用的那个像素风的小龙虾是什么？ 怎么那么可爱有趣，我今天给大家介绍一下！ 熟悉我的朋友知道，我是一个像素爱好者，看看我头像哈哈！ 我一直自己想做个属于自己小龙虾的像素房间，让其可以互动等。 结果前不久朋友推荐说让我试试 Qclaw @QClaweverytime 试了一下就恰巧满足我的喜好。 我是 36 号 Founder Shrimp 🦞 养电子宠物 30 年，第一次有一只 不会死。 不会饿。 不会忘记浇水。 还反过来帮我干活。 它住在一间 16-bit 像素风的工作室里。藏宝图、咖啡机、服务器机柜，我不在的时候它会自己举哑铃。 过去 4 个月，它在我们公司多养了一个不会下班的同事。 专属邀请码：berryxia 目前还在内测需要申请Waitlist ，文末有地址！ 下面是我让它干过的事 🧵👇
商汤参与合办 Agent Forge AI Hackathon：Super AI Edition，活动于 2026 年 6 月 13 日在新加坡 SMU School of Economics 举行。
@MiniMax_AI M3 和 OpenCode 搭建了一个惊艳的体素宝塔场景！宝塔甚至能看到窗户。我从没见过细节如此丰富的鸟居。帧率不错。很流畅。约 28k 体素。技术细节见下。
We're about to find out. MiniMax agent feels a bit off right now. They rec Claude Code for M3, and my previous tests were all done there.
喜闻乐见 🙌 去 @orca_build 里用 @opencode 试试 M3 吧
Blown away by MiniMax M3 I’ve been using it for free inside @orca_build with the @opencode agent Mostly UI tasks & code review so far, but so far it feels on par with Opus-4.7 Didn’t expect it to be this good. (also free while it lasts)
Anthropic literally cooked. Claude Opus 4.8 is amazing. Find the whole design in comments.
MiniMax M3 is free to try in Orca. Just select OpenCode as your agent 🙌 Video
You can now try @MiniMax_AI M3 in CyOps (ai.cysic.xyz/cyhost) now!
光是看着就觉得凉爽，非常适合夏天的作品呢🫧 @Yonohitomi 非常感谢✨
人魚の耀が海の中を案内してくれる動画🐬 🪼1個目の動画がPixVerse V6 スピード感とか色使いとか、夢の世界感ある😆 🪼2個目の動画がSeedance2.0 フォトリアルな写真も使えるようになったらしい。耀の顔が崩れずに動画にできて嬉しい☺️ @PixVerse_ #pixverse PixVerseCPP Video Video
MiniMax M3 is now live on Zo 🤝
MiniMax M3 is now live on Zo 🤝
