July 10 news.OpenAI The GPT-5.6 series, comprising the flagship Sol, the equilibrium Terra and the value for money Luna model, was officially published。
The official claim that Sol scored 53.6, 13.1 points higher than Claude Fable 5 in a long-range smart body assessment across 55 industries; in programming, Sol scored 80 points on the Artificial Analysis programming smart body index, refreshing the record, exporting token less than half of Fable 5, at a cost of about one third。
Designed to enhance judgement, aesthetic interfaces can be generated in high-level direction and, with enhanced computer-operated capability, to check the rendering, detect and correct problems and deliver them
THE KNOWLEDGE OFFICE LANDSCAPE HAS IMPROVED SIGNIFICANTLY, ALLOWING FOR A MORE ACCURATE EXTRAPOLATION OF THE DESIGN SYSTEM OF THE REFERENCE DOCUMENT AND ITS APPLICATION TO NEW CONTENT, AND THE REFINEMENT OF FORMULAE AND LAYOUT QUALITY OF PPT, DOCUMENTS AND SPREADSHEETS
Add a new ultra reasoning slot, with 4 AI smart bodies working in parallel by default, up to 16, suitable for high-intensity tasks
There has been a significant increase in network security capabilities, with the Expluit Bench rating rising from the previous generation of 47.9% to 73.5%。
GPT-5.6 Add a new Programme Tool Calling, with the model preparing its own lightweight program coordination tool to support zero data retention (ZDR). The price was: Sol, $5 per million token, $30 per output; Terra, $2.5 and $15; Luna, $1 and $6。
OpenAI also announced Codex App Merge ChatGPT Desktop applications, the Chat, Work and Codex modes are integrated into the same entry and the original ChatGPT desktop application is renamed ChatGPT Classic。