Up to four models
Gemini 4 Argon vs GPT-6.1 Sol
On the index (GPT-4 = 100), Gemini 4 Argon scores 191 and GPT-6.1 Sol 191. 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
GPT-6.1 Sol
191175–207
WeightsClosed
ReleasedSep 2026
Price in / out$2.0 / $10
Context1.05M
Core tests taken6 of 9
Gemini 4 Argon and GPT-6.1 Sol are too close to call (1 point 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 | GPT-6.1 Sol |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 96.4% | 95.4% |
| EBR-bench | 64.8% | 54.3% |
| SimpleBench | 91.7% | 82.9% |
| SimpleQA Verified | 76.0% | 73.9% |
| Maths | ||
| FrontierMath (tiers 1-3) | 86.0% | 93.7% |
| AIME-style maths (OTIS mock) | 100.0% | 100.0% |
| ProofBench v1.1 | 99.0% | 99.0% |
| Chess puzzles | 59.0% | 68.0% |
| FrontierMath (tier 4) | 78.0% | 100.0% |
| Coding | ||
| Terminal-Bench 4.0 | 57.6% | 55.0% |
| Code Migration | 68.2% | 65.1% |
| CyberBench v1.1 | 77.9% | 39.3% |
| IOI | 100.0% | 96.9% |
| ProgramBench | 2.5% | 3.0% |
| Vibe Code Bench v1.1 | 91.9% | 88.9% |
| MirrorCode | 58.9% | 61.1% |
| Agents | ||
| APEX-Agents | 82.4% | 60.0% |
| BioMysteryBench | 76.3% | 79.6% |
| SRE Bench | 44.3% | 50.8% |
| Terminal-Bench-Science | 44.3% | 52.9% |
| DTBench | 97.6% | 96.5% |
| Furniture assembly | 72.5% | 80.0% |
| LMCA | 63.7% | 62.8% |
| Mystery game puzzles | 64.0% | 80.0% |
| Professional work | ||
| Finance Agent (v2) | 65.4% | 52.0% |
| EMB | 75.2% | 70.8% |
| HLAB-Vals-1.1 | 27.1% | 6.3% |
| Legal Research Bench | 54.8% | 38.5% |
| MedCode | 58.8% | 48.8% |
| MedScribe | 87.4% | 86.5% |
| Public Benefits Bench v1.1 | 69.8% | 59.3% |
| SAGE | 89.5% | 90.5% |
| Tax Agent Bench | 49.2% | 21.8% |
| Novel problems | ||
| ARC-AGI-2 | 91.3% | 94.2% |
| ARC-AGI-1 | 98.5% | 98.5% |
| ARC-AGI-3 | 49.8% | 52.7% |
| MysteryMechanism | 45.5% | 46.4% |
Tests at least two of these models have taken. Bold rows are the core tests.