Up to four models
Claude Opus 5.5 vs GPT-6 Astra
On the index (GPT-4 = 100), Claude Opus 5.5 scores 195 and GPT-6 Astra 193. 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
195178–211
WeightsClosed
ReleasedSep 2026
Price in / out$4.0 / $20
Context1M
Core tests taken6 of 9
GPT-6 Astra
193178–209
WeightsClosed
ReleasedSep 2026
Price in / out$10 / $50
Context1.05M
Core tests taken7 of 9
Claude Opus 5.5 and GPT-6 Astra are too close to call (1 point apart, within the uncertainty), ahead on 3 of 6 shared core tests, and costs 2.5× less.
Test by testthird-party results
Claude Opus 5.5GPT-6 Astra
Every result as a table
| Test | Claude Opus 5.5 | GPT-6 Astra |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 94.1% | 95.8% |
| LiveBench Data Analysis | 80.3% | 83.0% |
| LiveBench IF | 67.0% | 75.6% |
| LiveBench Language | 86.3% | 89.4% |
| LiveBench Reasoning | 92.2% | 92.7% |
| EBR-bench | 71.4% | 76.2% |
| SimpleBench | 88.4% | 83.6% |
| SimpleQA Verified | 72.2% | 75.6% |
| Maths | ||
| FrontierMath (tiers 1-3) | 91.2% | 93.7% |
| AIME-style maths (OTIS mock) | 100.0% | 100.0% |
| LiveBench Mathematics | 97.1% | 96.8% |
| ProofBench v1.1 | 100.0% | 99.0% |
| Chess puzzles | 67.0% | 74.0% |
| FrontierMath (tier 4) | 95.0% | 97.6% |
| Coding | ||
| Terminal-Bench 4.0 | 65.1% | 59.6% |
| LiveBench Agentic Coding | 71.7% | 57.3% |
| LiveBench Coding | 89.3% | 80.4% |
| Code Migration | 66.6% | 67.7% |
| CyberBench v1.1 | 74.6% | 41.1% |
| IOI | 95.1% | 100.0% |
| ProgramBench | 18.5% | 5.5% |
| Vibe Code Bench 1-100 | 30.4% | 27.6% |
| Vibe Code Bench v1.1 | 90.3% | 89.6% |
| FrontierSWE | 62.3% | 65.5% |
| MirrorCode | 95.6% | 60.0% |
| Agents | ||
| APEX-Agents | 73.5% | 64.7% |
| BioMysteryBench | 79.3% | 79.3% |
| SRE Bench | 33.6% | 56.9% |
| Terminal-Bench-Science | 47.1% | 62.9% |
| DTBench | 98.9% | 97.3% |
| Furniture assembly | 83.3% | 80.0% |
| LMCA | 68.2% | 64.4% |
| Mystery game puzzles | 71.0% | 84.0% |
| Professional work | ||
| Finance Agent (v2) | 58.6% | 53.5% |
| EMB | 75.9% | 71.7% |
| HLAB-Vals-1.1 | 9.2% | 7.1% |
| Legal Research Bench | 50.5% | 39.4% |
| MedCode | 49.8% | 48.5% |
| MedScribe | 91.4% | 87.9% |
| Tax Agent Bench | 44.5% | 20.7% |
| Novel problems | ||
| ARC-AGI-2 | 93.3% | 95.0% |
| ARC-AGI-1 | 98.5% | 98.5% |
| MysteryMechanism | 49.5% | 53.1% |
Tests at least two of these models have taken. Bold rows are the core tests.