Woofun AI reports that OpenAI has launched the GPT-5.6 series, comprising the flagship Sol, balanced Terra, and budget Luna models, now available on ChatGPT, Codex, and API. In the Agents’ Last Exam benchmark, Sol achieved a score of 53.6, exceeding Claude Fable 5 by 13.1 points, while maintaining an 11.4-point lead under medium reasoning intensity at roughly one-quarter of the competitor's cost.

Sol also demonstrated superior programming capabilities, scoring 80 on the Artificial Analysis index—2.8 points above Fable 5—while using less than half the tokens and time at one-third the cost. The release introduces an ultra mode enabling four parallel agents for enhanced results and speed, with API pricing set at $5/$30 per million tokens for Sol, $2.5/$15 for Terra, and $1/$6 for Luna.