Up to four models
GPT-5.6 Sol vs GPT-6 Sol
On the index (GPT-4 = 100), GPT-5.6 Sol scores 179 and GPT-6 Sol 178. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
GPT-5.6 Sol
179164–195
WeightsClosed
ReleasedJul 2026
Price in / out$4.0 / $20
Context1.05M
Core tests taken5 of 9
GPT-6 Sol
178162–194
WeightsClosed
ReleasedSep 2026
Price in / out$2.0 / $10
Context1.05M
Core tests taken6 of 9
GPT-5.6 Sol and GPT-6 Sol are too close to call (2 points apart, within the uncertainty), ahead on 2 of 5 shared core tests, at 2.0× the price.
Test by testthird-party results
GPT-5.6 SolGPT-6 Sol
Every result as a table
| Test | GPT-5.6 Sol | GPT-6 Sol |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 93.5% | 94.3% |
| LiveBench Data Analysis | 79.8% | 81.2% |
| LiveBench IF | 71.9% | 68.6% |
| LiveBench Language | 87.7% | 85.3% |
| LiveBench Reasoning | 91.6% | 88.6% |
| EBR-bench | 44.8% | 53.3% |
| SimpleBench | 64.8% | 73.1% |
| SimpleQA Verified | 69.7% | 60.7% |
| Maths | ||
| FrontierMath (tiers 1-3) | 89.1% | 89.8% |
| AIME-style maths (OTIS mock) | 100.0% | 100.0% |
| LiveBench Mathematics | 96.2% | 96.4% |
| ProofBench v1.1 | 83.0% | 83.0% |
| Chess puzzles | 59.0% | 56.0% |
| FrontierMath (tier 4) | 82.9% | 90.0% |
| Coding | ||
| Terminal-Bench 4.0 | 37.9% | 44.4% |
| LiveBench Agentic Coding | 56.2% | 52.9% |
| LiveBench Coding | 83.9% | 81.8% |
| Code Migration | 52.9% | 57.2% |
| CyberBench v1.1 | 76.3% | 78.0% |
| IOI | 91.2% | 82.6% |
| ProgramBench | 1.5% | 2.0% |
| Vibe Code Bench v1.1 | 80.5% | 87.8% |
| MirrorCode | 20.0% | 48.9% |
| Agents | ||
| BioMysteryBench | 71.1% | 74.8% |
| SRE Bench | 30.5% | 30.5% |
| Terminal-Bench-Science | 20.0% | 30.0% |
| DTBench | 95.7% | 97.3% |
| Furniture assembly | 56.7% | 58.3% |
| LMCA | 58.5% | 59.1% |
| Mystery game puzzles | 58.0% | 56.0% |
| Professional work | ||
| Finance Agent (v2) | 53.8% | 49.0% |
| EMB | 72.3% | 71.5% |
| Legal Research Bench | 48.1% | 28.8% |
| MedCode | 44.0% | 47.1% |
| MedScribe | 85.2% | 82.0% |
| Public Benefits Bench v1.1 | 66.5% | 56.6% |
| Tax Agent Bench | 31.1% | 15.0% |
| Novel problems | ||
| ARC-AGI-2 | 92.5% | 89.6% |
| ARC-AGI-1 | 97.5% | 95.5% |
| ARC-AGI-3 | 7.8% | 4.6% |
| MysteryMechanism | 33.3% | 30.2% |
Tests at least two of these models have taken. Bold rows are the core tests.