GPT-5.6 Sol
OpenAI·
LLMsflagshipcloud
Overview
Preview of OpenAI's GPT-5.6 frontier model (Sol tier), part of a Sol/Terra/Luna family in limited preview to trusted partners via Codex and the API. Sol priced at $5/M input, $30/M output tokens. Not yet generally available.
Capabilities and innovations
Frontier reasoningAgentic codingTool use / function callingComputer useCybersecurity tasksSol / Terra / Luna model familyLimited preview (Codex + API)Token-efficient agentic workflows
Benchmarks
| Benchmark | Score | Source |
|---|---|---|
| GPQA Diamond | 95.2% | independent vals.ai independent evaluation (accessed 2026-07-25). |
| Humanity’s Last Exam (no tools) | 47.2% | third party HLE 47,2% no-tools, measured by Artificial Analysis (max effort; xhigh gives 44,7%). |
| SWE-bench Pro | 64.6% | vendor 64.6% SWE-bench Pro. OpenAI separately reports that ~30% of SWE-bench Pro tasks are broken and advises caution on the benchmark. |
| MMLU-Pro | 89.1% | independent vals.ai independent evaluation (accessed 2026-07-25). |
| Terminal-Bench 2.x | 85.77% | independent Terminal-Bench 2.1 85.77%, measured by Vals AI with the Terminus 2 harness (accessed 2026-09-23). Vals AI is the reference harness for this field so values compare across vendors; vendor and other third-party figures stay in the notes below. Replaces 88.8 (vendor). |
| Terminal-Bench 3.0 | 34.6% | independent Terminal-Bench 3.0 34.6% (± 3.1, 95% CI), official leaderboard maintained by the Terminal-Bench team (Stanford / Harbor / Laude Institute), max effort, agent harness: Codex (accessed 2026-09-23). Do not compare against the terminal_bench column: 3.0 is a harder generation on a different scale. |
| Terminal-Bench 4.0 | 37.3% | independent Terminal-Bench 4.0 37.3% (± 3.8, 95% CI), official leaderboard maintained by the Terminal-Bench team (Stanford / Harbor / Laude Institute), max effort, agent harness: Codex (accessed 2026-09-23). 4.0 is its own generation and not comparable with the 2.x or 3.0 fields. |
| MMMU-Pro | 83.4% | third party MMMU-Pro 83,4%, measured by Artificial Analysis (max effort). |
Reliability
- Hallucination rate (Vectara HHEM)
- 12.4% (lower is better)independent
- tau3_banking
- 44.3 ()third party
Links
More from OpenAI
Data curated by AI Model Timeline. See the methodology for admission criteria, benchmark eras and source priority.