Annonces

Nvidia's Nemotron-4, the speed-focused Celeris-1, and a pathogen surveillance benchmark hitting a 50% ceiling reveal that current AI benchmarks fail to capture

Chinese AI labs Qwen, DeepSeek, MiniMax, and Kimi have matched or beaten US frontier models in reasoning, coding, and document parsing, using pricing and open-w

Moonshot AI releases Kimi K3, a 2.8T-parameter open-source model with 1M-token context, trailing Claude Fable 5 and GPT 5.6 Sol on most benchmarks but beating t