Blog

Xiaohongshu (RedNote) launches dots.llm1, a 142B parameter MoE model challenging frontier AI standards with open weights.

Mistral AI releases Magistral Small, a 24B parameter reasoning model with extended thinking capabilities, open-sourced under Apache 2.0.

Google DeepMind releases Gemini 2.5 Pro Preview 06-05 with 1M token context, enhanced reasoning, and agentic capabilities. A technical breakdown for developers.

MiniMax introduces MiniMax-M1, a large-scale open-source model featuring lightning attention architecture and a 1M token context window, challenging top-tier commercial systems.

DeepSeek's mid-cycle R1 update delivers massive benchmark leaps, improved coding capabilities, and native function calling for developers.

Anthropic releases Claude Sonnet 4 on May 22, 2025, featuring a 200K context window, native tool calling, and optimized performance for complex agent workflows.

Anthropic launches Claude Opus 4 on May 22, 2025, marking a historic leap in autonomous reasoning with 200K context and parallel tool use.

Mistral AI releases Devstral, an open-source 24B parameter model optimized for software engineering and agentic coding tasks under Apache 2.0.

TII releases Falcon H1 with hybrid SSM+attention, 0.5B-34B sizes, Apache 2.0 license, and top benchmark scores.