Up to four models
Claude Opus 5.5 vs Gemini 4 Argon
On the index (GPT-4 = 100), Claude Opus 5.5 scores 195 and Gemini 4 Argon 191. 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
191176–207
WeightsClosed
ReleasedSep 2026
Price in / out$2.0 / $10
Context1.05M
Core tests taken6 of 9
Claude Opus 5.5 and Gemini 4 Argon are too close to call (3 points apart, within the uncertainty), ahead on 3 of 6 shared core tests, at 2.0× the price.
Test by testthird-party results
Claude Opus 5.5
Gemini 4 Argon
Every result as a table
| Test | Claude Opus 5.5 | Gemini 4 Argon |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 94.1% | 96.4% |
| EBR-bench | 71.4% | 64.8% |
| SimpleBench | 88.4% | 91.7% |
| SimpleQA Verified | 72.2% | 76.0% |
| Maths | ||
| FrontierMath (tiers 1-3) | 91.2% | 86.0% |
| AIME-style maths (OTIS mock) | 100.0% | 100.0% |
| ProofBench v1.1 | 100.0% | 99.0% |
| Chess puzzles | 67.0% | 59.0% |
| FrontierMath (tier 4) | 95.0% | 78.0% |
| Coding | ||
| Terminal-Bench 4.0 | 65.1% | 57.6% |
| Code Migration | 66.6% | 68.2% |
| CyberBench v1.1 | 74.6% | 77.9% |
| IOI | 95.1% | 100.0% |
| ProgramBench | 18.5% | 2.5% |
| Vibe Code Bench 1-100 | 30.4% | 21.2% |
| Vibe Code Bench v1.1 | 90.3% | 91.9% |
| FrontierSWE | 62.3% | 55.0% |
| MirrorCode | 95.6% | 58.9% |
| Agents | ||
| APEX-Agents | 73.5% | 82.4% |
| BioMysteryBench | 79.3% | 76.3% |
| SRE Bench | 33.6% | 44.3% |
| Terminal-Bench-Science | 47.1% | 44.3% |
| DTBench | 98.9% | 97.6% |
| Furniture assembly | 83.3% | 72.5% |
| LMCA | 68.2% | 63.7% |
| Mystery game puzzles | 71.0% | 64.0% |
| Professional work | ||
| Finance Agent (v2) | 58.6% | 65.4% |
| EMB | 75.9% | 75.2% |
| HLAB-Vals-1.1 | 9.2% | 27.1% |
| Legal Research Bench | 50.5% | 54.8% |
| MedCode | 49.8% | 58.8% |
| MedScribe | 91.4% | 87.4% |
| Public Benefits Bench v1.1 | 70.6% | 69.8% |
| SAGE | 91.3% | 89.5% |
| Tax Agent Bench | 44.5% | 49.2% |
| Novel problems | ||
| ARC-AGI-2 | 93.3% | 91.3% |
| ARC-AGI-1 | 98.5% | 98.5% |
| MysteryMechanism | 49.5% | 45.5% |
Tests at least two of these models have taken. Bold rows are the core tests.