This AINews digest highlights a sharp increase in OpenAI's agent product adoption alongside significant advances in local inference and multimodal AI capabilities.

  • OpenAI reports that usage of Codex + ChatGPT Work grew 2.5x in a week, with GPT-5.6 Sol demand described as "insane" by CEO Sam Altman.
  • PrismML released Bonsai 27B, a compact model based on Qwen 3.6 27B, available in 1-bit and ternary variants for local deployment.
  • Tencent Hunyuan released Hy3, a 295B flagship-scale model that can be served on a single GPU using llama.cpp with MTP enabled.
  • OpenMOSS launched MOSS-VL-Realtime, an 11B vision-language model designed for continuous video streams and active evidence search.
  • Perplexity open-sourced WANDR, a 500-task benchmark for wide-and-deep agentic research that re-fetches cited pages to check claims against underlying evidence.

The coverage suggests a shift in the AI landscape where high-performance agent tools are seeing rapid enterprise adoption while open-source compression techniques make frontier-level capabilities viable on consumer hardware.