Moonshot AI publie Kimi K3 : 2,8 trillions de paramètres en open weight. 7× plus gros que Llama 3.1 405B, et ça tient sur 4 RTX 4090 grâce à KTransformers, leur couche d’inférence optimisée publiée en #opensource.Ce qui est intéressant : contexte 1M tokens, coût annoncé à 10-15 % de GPT-5.5 pour des performances comparables à Claude Opus 4.8. Distribution le 27 juillet.https://tit4nium.fr/blog/kimi-k3-plus-gros-modele-open-weight/#llm #ia
Related
Unlimited AI tokens aren’t unlimited after all as US Army burns through supplyhttps://arstechnica.com/ai/2026/07/us-army...
Unlimited AI tokens aren’t unlimited after all as US Army burns through supplyhttps://arstechnica.com/ai/2026/07/us-army-faces-ai-use-limits-after-exhausting-years-supply-of-ai-tok...
📰 The Best Backpacking Sleeping Pads, Tested on the Trail (2026)Our top-pick sleeping pads from Nemo, Therm-a-Rest, and ...
📰 The Best Backpacking Sleeping Pads, Tested on the Trail (2026)Our top-pick sleeping pads from Nemo, Therm-a-Rest, and Gossamer Gear use high-tech materials to engineer the best s...
Is Context7 the best solution for vibe-coding to ensure you're always using the latest version of a component / library ...
Is Context7 the best solution for vibe-coding to ensure you're always using the latest version of a component / library ?#mcp #llm #dev #ai