GPT-5.6 Sol
Jul 2026 · Proprietary · Current · supported evidence
OpenAI’s flagship. Leads ARC-AGI-2 and Terminal-Bench. Token-efficient per task even though list price looks high.
Intelligence
60.9
AA Index
API price
$5 / $30
Input / output per 1M
Context
1.05M
74 tok/s
Capability
Intelligence Index60.9
Coding index77.4
Agentic index57.8
Arena Elo (offset)182
Use it when
- ▸Abstract reasoning
- ▸Codex-style agents
- ▸Tool-heavy apps
Skip if
- –You need the cheapest intelligence point
Benchmarks
GPQA Diamond Graduate-level science questions designed so Google search is not enough. Still one of the cleanest knowledge/reasoning splits. | 94.6% |
|---|---|
Humanity’s Last Exam Expert-written questions across fields. Harder than MMLU; the current differentiator for “does this model actually know things.” | 49.5% |
MMLU-Pro Harder, less-saturated successor to MMLU. Classic MMLU is above 90% for every flagship and no longer ranks the field. | — |
SWE-bench Verified 500 human-validated GitHub issues. Score swings 5–15 points by harness — treat vendor numbers as an upper bound. | — |
SWE-bench Pro Harder, contamination-resistant coding eval. Currently the best public split between “can code” and “can maintain a repo.” | 64.6% |
Terminal-Bench 2.1 End-to-end tasks in a real terminal. Better proxy for coding agents than HumanEval, which is fully saturated. | 88.8% |
ARC-AGI-2 Abstract visual puzzles. Rewards generalization over memorization. GPT-5.6 Sol currently leads the published set. | 92.5% |
AIME 2025 American Invitational Mathematics Examination. Contest math; reasoning-mode models dominate. | — |
MMMU College-level multimodal understanding across diagrams, charts, and exam figures. | — |
List price cut in late August 2026; some boards still show $4/$20.
Modalities
text · vision · audio · tools
Reasoning mode
reasoning
Hybrid and reasoning models spend tokens thinking. That raises GPQA and agents, and also raises latency and bill.