Tencent Hy4 Ranks #2 on ChessQA, Solves 34 of 35 Puzzles
Tencent's Hy4 model placed second among all models on ChessQA, correctly solving 34 of 35 chess puzzles by taking time to reason rather than blitzing.
Tencent's Hy4 model placed second among all models on ChessQA, correctly solving 34 of 35 chess puzzles by taking time to reason rather than blitzing.
OpenCode reports that 44% of its tokens, about 16T, were served by its own GPUs, with 10x more hardware coming online in the next few months.
OpenCode reports that 44% of its daily token traffic is now served by its own GPUs, totaling 16 trillion tokens, with 10x more hardware coming online in the nex…
Cursor added Gemini 4 Neon, which scores 63.7% on CursorBench 4.1 at default effort, up from 57.3% for Gemini 4 Hydrogen, priced at $3/M input and $15/M output…
Command Code has added Gemini 4 Neon, featuring a 1M token context window and availability across Pro, Max, and API tiers.
OpenAI shares new data showing researchers' AI use is rising exponentially, agents handle longer tasks, and human involvement remains central, with safety and a…
OpenAI shares a recap of the process for moving a frontier model from training to internal use to public release, outlining four layers of protection.
OpenAI shares how Codex can use existing developer tools while users control what it can access, explaining the sandbox architecture behind it.
LiteLLM announced a startup program starting with Y Combinator companies and open to other VC-backed AI startups with 5+ people, bundling its Rust AI Gateway, a…
OpenAI's Tibo says GPT-6.2 Astra performed well on hard tasks, citing team results showing up to 14% PPA improvement on AMD flagship GPU IP in under five months…