GPT-5.6 Officially Released: Agent Evaluates Leading Claude Fable Model at 13.1 Points

According to Vision One monitoring, OpenAI has released the GPT-5.6 series, which includes the flagship model Sol, the balanced version Terra, and the budget version Luna. The new models are now live on ChatGPT, Codex, and the API.

GPT-5.6 Sol focuses on performance and efficiency. In a professional Agent evaluation called Agents’ Last Exam, Sol scored 53.6, surpassing Claude Fable 5 by 13.1 points. Even under medium reasoning intensity, it still leads by 11.4 points, with an estimated cost of about a quarter of the competitor.

There has also been an improvement in programming capability. Sol scored 80 in the Artificial Analysis programming Agent index, 2.8 points higher than Fable 5, while outputting less than half the tokens, taking less than half the time, and costing about a third less.

GPT-5.6 has introduced an ultra mode as well. By default, it allows 4 Agents to work in parallel, exchanging more tokens for stronger results and shorter processing time. API users can also access a multi-Agent test version.

For every 1 million tokens, the input and output prices for the API are: Sol $5 and $30, Terra $2.5 and $15, Luna $1 and $6.