Up to four models
Gemini 4 Argon vs Claude Sonnet 5.5
On the index (GPT-4 = 100), Gemini 4 Argon scores 191 and Claude Sonnet 5.5 187. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
191176–207
WeightsClosed
ReleasedSep 2026
Price in / out$2.0 / $10
Context1.05M
Core tests taken6 of 9
Claude Sonnet 5.5
187171–203
WeightsClosed
ReleasedSep 2026
Price in / out$2.0 / $10
Context1M
Core tests taken7 of 9
Gemini 4 Argon and Claude Sonnet 5.5 are too close to call (5 points apart, within the uncertainty), ahead on 4 of 6 shared core tests, at a similar price.
Test by testthird-party results
Every result as a table
| Test | Gemini 4 Argon | Claude Sonnet 5.5 |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 96.4% | 95.6% |
| SimpleBench | 91.7% | 75.9% |
| SimpleQA Verified | 76.0% | 46.5% |
| Maths | ||
| FrontierMath (tiers 1-3) | 86.0% | 88.8% |
| AIME-style maths (OTIS mock) | 100.0% | 100.0% |
| ProofBench v1.1 | 99.0% | 100.0% |
| Chess puzzles | 59.0% | 59.0% |
| FrontierMath (tier 4) | 78.0% | 80.5% |
| Coding | ||
| Terminal-Bench 4.0 | 57.6% | 64.1% |
| Code Migration | 68.2% | 69.8% |
| CyberBench v1.1 | 77.9% | 59.6% |
| IOI | 100.0% | 83.1% |
| ProgramBench | 2.5% | 6.5% |
| Vibe Code Bench v1.1 | 91.9% | 92.4% |
| FrontierSWE | 55.0% | 61.9% |
| MirrorCode | 58.9% | 97.8% |
| Agents | ||
| APEX-Agents | 82.4% | 75.5% |
| BioMysteryBench | 76.3% | 81.1% |
| SRE Bench | 44.3% | 30.1% |
| Terminal-Bench-Science | 44.3% | 47.1% |
| DTBench | 97.6% | 97.3% |
| Furniture assembly | 72.5% | 75.0% |
| LMCA | 63.7% | 68.3% |
| Mystery game puzzles | 64.0% | 65.0% |
| Professional work | ||
| Finance Agent (v2) | 65.4% | 58.1% |
| EMB | 75.2% | 75.7% |
| HLAB-Vals-1.1 | 27.1% | 10.4% |
| Legal Research Bench | 54.8% | 48.1% |
| MedCode | 58.8% | 52.9% |
| MedScribe | 87.4% | 91.1% |
| Public Benefits Bench v1.1 | 69.8% | 67.2% |
| SAGE | 89.5% | 88.0% |
| Tax Agent Bench | 49.2% | 42.0% |
| Novel problems | ||
| ARC-AGI-2 | 91.3% | 83.9% |
| ARC-AGI-1 | 98.5% | 96.5% |
| ARC-AGI-3 | 49.8% | 24.3% |
| MysteryMechanism | 45.5% | 49.1% |
Tests at least two of these models have taken. Bold rows are the core tests.