Étiquette : china

MiniMax M2.5 scores 80.2% on SWE-Bench Verified, runs at 100 tokens/sec, and costs $1/hour. Here’s what we know — and what’s still unverified.

En septembre 2025, un groupe de pirates informatiques parrainé par l'État chinois aurait utilisé Claude Code (un assistant de programmation d'IA créé par Anthropic) pour mener une opération d'espionnage à grande échelle contre une trentaine de cibles mondiales, dont de grandes entreprises technologiques, des institutions financières, des fabricants de produits chimiques et des agences gouvernementales. Contrairement aux attaques précédentes assistées par l'IA, les attaquants ont construit un cadre automatisé autour de Claude Code afin que [...]

Le 6 novembre 2025, Moonshot AI, soutenu par Alibaba, a publié Kimi K2 Thinking, son modèle open-source le plus avancé à ce jour. Il s'agit de la première variante axée sur le raisonnement de la famille Kimi K2, qui marque une avancée majeure dans le domaine du raisonnement en contexte long et en plusieurs étapes, ainsi que dans l'utilisation autonome d'outils. Kimi K2 Thinking a immédiatement fait les gros titres pour ses performances : il a atteint de nouveaux scores sur plusieurs benchmarks ouverts, dont Humanity's Last [...]

Alibaba’s AI stack just crossed a symbolic frontier. On September 5 2025, the company’s cloud division unveiled Qwen-3-Max-Preview (Instruct), a 1-trillion-parameter large-language model that immediately takes its place among the most complex systems ever deployed and the most powerful AI models. The rollout began quietly inside Qwen Chat and Alibaba Cloud Model Studio. Soon coming to Fello AI. Only […]

Moonshot AI, Chinese startup backed by Alibaba and Tencent, is rolling out an upgraded version of its open‑weight open-source large language model, Kimi K2. The new build Kimi K2‑0905, introduces a 256,000-token context window, improved coding performance, and retains its permissive modified MIT license. Although early beta access was briefly mentioned on the company’s Discord […]

The open-source AI race just got more interesting. Chinese startup DeepSeek has unveiled DeepSeek-V3.1, its biggest upgrade yet, bringing sharper reasoning, stronger coding skills, and new support for tool-calling and agent workflows. Unlike its earlier release that felt experimental, V3.1 arrives as a serious contender. With a 128K context window, hybrid reasoning modes, and an API that’s dramatically cheaper […]

A small Chinese startup has just released a text-to-video model that’s beating Google’s much-hyped Veo 3 across nearly every major benchmark. Meet MiniMax’s Hailuo 02. From physics simulation to prompt accuracy, Hailuo 02 is winning over creators, developers, and the internet at large. And perhaps the most remarkable part? It’s available to everyone for just $8 a month, compared to Veo’s top-tier […]

In late May 2025, Chinese startup DeepSeek quietly rolled out R1-0528, a beefed-up version of its open-source R1 reasoning model. The upgrade boosts parameter count from 671 billion to 685 billion, adds a lightweight distilled variant, and publishes all weights, training recipes and docs under an MIT license on GitHub and Hugging Face. Under the […]

Just when you thought Silicon Valley had the AI game locked down, Alibaba has unleashed Qwen3, a new generation of AI models so powerful they’re making US tech giants sweat. This isn’t just another AI; it’s a statement – a declaration that China is not just catching up, but is ready to lead the AI […]

The internet is ablaze with footage of what looks like a cutting-edge AI robot going completely berserk, flailing wildly and seemingly lunging at terrified engineers. Is this the moment we all feared? Are the machines finally fighting back, or is there a more innocent explanation for this metallic meltdown? Buckle up, because the truth might […]