Thinking Machines released Inkling-Small, a 276B-parameter mixture-of-experts model with 12B active parameters that retains multimodal reasoning and a 1M-token context window while using less compute. OpenAI reduced GPT-5.6 Luna pricing by 80% and Terra pricing by 20%, while improving Sol's API speed across API, Codex, and ChatGPT Work subscriptions.

  • Inkling-Small uses substantially less compute than previous iterations while maintaining variable thinking effort.
  • OpenAI extended efficiency gains from the price cuts to its broader suite of developer tools and subscriptions.
  • Gemini Robotics ER 2 enhances automation capabilities through advanced AI and LLM integration for precise industrial tasks.
  • Open-weight models like GLM 5.2 and Kimi K3 reached accuracy parity with proprietary models on the ClinReg benchmark at one-third the cost.

These updates lower infrastructure costs for developers and improve the economic viability of open-source models in regulated industries.