Up to four models
GPT-6 vs Claude Opus 5.5
On the index (GPT-4 = 100), GPT-6 scores 199 and Claude Opus 5.5 197. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
Claude Opus 5.5
197181–214
WeightsClosed
ReleasedSep 2026
Price in / out$4.0 / $20
Context1M
Core tests taken6 of 9
GPT-6 and Claude Opus 5.5 are too close to call (1 point apart, within the uncertainty), ahead on 3 of 6 shared core tests.
Test by testthird-party results
GPT-6Claude Opus 5.5
Every result as a table
| Test | GPT-6 | Claude Opus 5.5 |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 95.8% | 90.6% |
| LiveBench Data Analysis | 83.0% | 80.3% |
| LiveBench IF | 75.6% | 67.0% |
| LiveBench Language | 89.4% | 86.3% |
| LiveBench Reasoning | 92.7% | 92.2% |
| EBR-bench | 76.2% | 71.4% |
| SimpleQA Verified | 75.6% | 72.2% |
| Maths | ||
| FrontierMath (tiers 1-3) | 93.7% | 91.2% |
| AIME-style maths (OTIS mock) | 100.0% | 100.0% |
| LiveBench Mathematics | 96.8% | 97.1% |
| ProofBench v1.1 | 99.0% | 100.0% |
| Chess puzzles | 74.0% | 67.0% |
| FrontierMath (tier 4) | 97.6% | 95.0% |
| Coding | ||
| Terminal-Bench 4.0 | 59.6% | 65.1% |
| LiveBench Agentic Coding | 57.3% | 71.7% |
| LiveBench Coding | 80.4% | 89.3% |
| Code Migration | 67.7% | 66.6% |
| CyberBench v1.1 | 41.1% | 74.6% |
| IOI | 100.0% | 95.1% |
| ProgramBench | 5.5% | 18.5% |
| Vibe Code Bench 1-100 | 27.6% | 30.4% |
| Vibe Code Bench v1.1 | 89.6% | 90.3% |
| FrontierSWE | 65.5% | 62.3% |
| Agents | ||
| APEX-Agents | 64.7% | 73.5% |
| BioMysteryBench | 79.3% | 79.3% |
| SRE Bench | 56.9% | 33.6% |
| Terminal-Bench-Science | 62.9% | 47.1% |
| DTBench | 97.3% | 98.9% |
| Furniture assembly | 80.0% | 83.3% |
| LMCA | 64.4% | 68.2% |
| Mystery game puzzles | 84.0% | 71.0% |
| Professional work | ||
| Finance Agent (v2) | 53.5% | 58.6% |
| EMB | 71.7% | 75.9% |
| HLAB-Vals-1.1 | 7.1% | 9.2% |
| Legal Research Bench | 39.4% | 50.5% |
| MedCode | 48.5% | 49.8% |
| MedScribe | 87.9% | 91.4% |
| SAGE | 46.4% | 45.8% |
| Tax Agent Bench | 20.7% | 45.6% |
| Novel problems | ||
| ARC-AGI-2 | 95.0% | 93.3% |
| ARC-AGI-1 | 98.5% | 98.5% |
| MysteryMechanism | 53.1% | 49.5% |
Tests at least two of these models have taken. Bold rows are the core tests.
