GPT-5.6: Frontier intelligence that scales with your ambition
OpenAI Blog · 2026-07-09
OpenAI launches GPT-5.6, a three-tier flagship model family (Sol, Terra, Luna), claiming state-of-the-art performance in coding, cybersecurity, and agentic workflows while outperforming Claude Fable 5 on most benchmarks at lower estimated cost.
Appears in
Extraction
Topics: large-language-modelsmodel-releaseai-benchmarksagentic-aiai-safety
Claims
- GPT-5.6 Sol scores 53.6 on Agents' Last Exam, outperforming Claude Fable 5 by 13.1 points at roughly one-quarter the estimated cost.
- GPT-5.6 Sol achieves 80 on the Artificial Analysis Coding Agent Index, 2.8 points above Fable 5, using less than half the output tokens and costing about one-third less.
- GPT-5.6 Sol scores 62.6% on OSWorld 2.0, surpassing Claude Opus 4.8 while using 85% fewer output tokens.
- GPT-5.6 introduces an 'ultra' mode that coordinates four parallel agents by default for demanding tasks, trading higher token use for stronger results and faster completion.
- GPT-5.6's cybersecurity capabilities score 73.5% on ExploitBench2 versus GPT-5.5's 47.9%, and the model does not cross OpenAI's Critical threshold in biology or cybersecurity risk.
Key quotes
We trained GPT‑5.6 to get more useful work from every token.
GPT‑5.6 Sol sets a new high of 53.6, eclipsing Claude Fable 5 (adaptive reasoning) by 13.1 points. Even at medium reasoning, it beats Fable 5 by 11.4 points at roughly one-quarter the estimated cost.
Compared with previous models, our GPT‑5.6 Sol cyber safeguards block roughly ten times more potentially harmful activity.