GPT-5.6 Sol vs GPT-6 Sol
Vals Science Index?
GPT-5.6 Sol41.5
GPT-6 Sol45
| Benchmark | GPT-5.6 Sol | GPT-6 Sol |
|---|---|---|
| Humanity's Last Exam | 42.2 | 41 |
| Arena Text · overall | 1,485 · xhigh | 1,459 · max |
| Arena Text · Chinese | 1,525.8 · xhigh | 1,412 · max |
| Arena Text · English | 1,457.8 · xhigh | 1,396.4 · max |
| Arena Text · French | 1,471.7 · xhigh | 1,394.7 · max |
| Arena Text · German | 1,474.9 · xhigh | 1,373.4 · max |
| Arena Text · instruction following | 1,482.7 · xhigh | 1,411.2 · max |
| Arena Text · Japanese | 1,473 · xhigh | 1,368.6 · max |
| Arena Text · Korean | 1,442.4 · xhigh | 1,333.5 · max |
| Arena Text · non-English | 1,450.9 · xhigh | 1,387.9 · max |
| Arena Text · Polish | 1,458 · xhigh | 1,379.7 · max |
| Arena Text · Russian | 1,471.7 · xhigh | 1,403.1 · max |
| Arena Text · Spanish | 1,442.4 · xhigh | 1,386.6 · max |
| AA Omniscience Index | 19.4 | 27 |
| SimpleQA Verified v1 | 69.7 · max | 64.4 |
| CritPt (physics) | 22.9 | 24.6 |
| AA Intelligence Index 4.3.2 | 39.2 | 39.8 |
| AA Omniscience · Accuracy | 57.8 | 53.5 |
| AA Omniscience · Hallucination Rate | 90.8 | 56.8 |
| AA-LCR (long context) 1.1 | 80.3 | 82.3 |
| GDP.pdf · AA all-pass | 26.2 | 23.8 |
| MLCR-AA | 26.1 | 28.9 |
| ARC-AGI-1 · public evaluation | 93.9 | 88.9 |
| ARC-AGI-1 · semi-private | 92.5 | 83.7 |
| ARC-AGI-2 · public evaluation | 70.7 | 49.4 |
| ARC-AGI-2 · semi-private | 67.1 | 57.8 |
| ARC-AGI-3 · semi-private | 1.1 | 0.4 |
| Chess Puzzles | 59 · max | 56 · max |
| FrontierMath Erdos | 0 · max | 1.5 · max |
| FrontierMath Tier 4 (v2) | 82.9 · max | 90 · max |
| FrontierMath Tiers 1–3 (v2) | 89.1 · max | 89.8 · max |
| Mystery Game Puzzles | 58 · max | 56 · max |
| MathArena · ArXivMath · June 2026 | 88.5 · max | 91 · max |
| MathArena · BrokenArXiv · June 2026 | 67.3 · max | 95.4 · max |
| Vals MedCode | 44 | 47.1 |
| Vals MedScribe | 85.2 | 82 |
| Vals Legal Research Bench | 48.1 | 28.9 |
| Vals Science Index | 41.5 | 45 |
| IFBench (instructions) | 69.6 | — |
| MathArena · ArXivMath · August 2026 | — | 93 · max |
| MathArena · ArXivMath · May 2026 | — | 90 · max |
| MathArena · BrokenArXiv · August 2026 | — | 78.6 · max |
| MathArena · BrokenArXiv · May 2026 | — | 87 · max |
| Vals LegalBench | 87 | — |
Specs
| Spec | GPT-5.6 Sol | GPT-6 Sol |
|---|---|---|
| Input, per 1M | $4 | $2 |
| Output, per 1M | $20 | $10 |
| Cost per task? | $0.50 | $0.25 |
| Speed, tok/s | 75 | — |
| Context | 1.05M | 1.05M |
| Weights | Closed | Closed |
| Released | Jul 2026 | Sep 2026 |
| See images | ✓ | ✓ |
| Read PDFs | Partial | Partial |
| Hear audio | — | — |
| Watch video | — | — |
| Call tools | ✓ | ✓ |
| Force a tool call | Partial | Partial |
| Return strict JSON | ✓ | ✓ |
| Search the web | ? | ? |
| Run code | ? | ? |
| Use a computer | ? | ? |
What breaks when you switch
GPT-5.6 Sol → GPT-6 Sol
Nothing known.
GPT-6 Sol → GPT-5.6 Sol
Nothing known.