OpenAI released GPT-5.6 in three sizes — Luna, Terra, and Sol — with pricing ranging from $1/$6 to $5/$30 per million input/output tokens. The models claim strong agentic performance, with Sol scoring 53.6 on the Agents' Last Exam benchmark, outperforming Claude Fable 5 by 13.1 points. Notably, OpenAI published a critique of SWE-Bench Pro the day before the release, estimating ~30% of its tasks are broken — a benchmark where Fable 5 significantly outperformed GPT-5.6 Sol (80% vs 64.6%). New API features include Programmatic Tool Calling (composing JavaScript to orchestrate tool calls), native multi-agent support, explicit prompt cache breakpoints, and an option to skip image resizing. The author notes GPT-5.6 Sol is competent but hasn't yet surpassed Fable 5 for complex coding tasks in personal use.
