Comparing 2 models
Gemini 3.7 Flash vs GPT-6.1 Sol
Benchmarks through
Verdict
Gemini 3.7 Flash and GPT-6.1 Sol split the 14 tasks where both are ranked, 7 each. Gemini 3.7 Flash costs 2.7x less per input token. GPT-6.1 Sol accepts 1.0x more context.
- Choose Gemini 3.7 Flash for Document Parsing (#2), OCR & Documents (#2), Vision & Documents (#2).
- Choose GPT-6.1 Sol for Image Understanding (#1), Translation (#2), Literary Translation (#2).
Built from the published task rankings below; ranks come from each task's own weighted benchmarks, so a win on one task says nothing about another.
Measured head to head
Benchmarks both models were run on
| Evaluation | Gemini 3.7 Flash | GPT-6.1 Sol | Source |
|---|---|---|---|
| Design Arena Coding | 1311 Elo | 1346 Elo | source · 2026-10-16 |
| Design Arena UI | 1301 Elo | 1331 Elo | source · 2026-10-16 |
| Design Arena Website | 1306 Elo | 1323 Elo | source · 2026-10-16 |
| FACTS Grounding | 75.5% | 82.4% | source · 2026-08-27 |
| FrontierMath Tiers 1-3 (Epoch AI) | 71.6% ± 2.7 | 93.7% ± 1.4 | source · 2026-08-14 |
| GPQA Diamond (Epoch AI) | 94.8% ± 1.3 | 95.4% ± 1.4 | source · 2026-08-14 |
| LiveBench Agentic Coding | 58.3% | 54.5% | source · 2026-10-16 |
| LiveBench Coding | 78.9% | 80.4% | source · 2026-10-16 |
| LiveBench Data Analysis | 68.0% | 82.7% | source · 2026-10-16 |
| LiveBench Instruction Following | 79.9% | 74.2% | source · 2026-10-16 |
| LiveBench Language | 85.5% | 90.1% | source · 2026-10-16 |
| LiveBench Math | 93.5% | 96.8% | source · 2026-10-16 |
| LiveBench Reasoning | 87.8% | 92.6% | source · 2026-10-16 |
| LMArena Agent | -0.4 | 14.8 | source · 2026-10-14 |
| LMArena Agent Bash Recovery | -5.4 | 4.5 | source · 2026-10-08 |
| LMArena Agent Steerability | -3.2 | 14.7 | source · 2026-10-14 |
| LMArena Agent Task Outcome | -0.8 | 15.7 | source · 2026-10-14 |
| LMArena Agent Tool Hallucination | 0.4 | 0.6 | source · 2026-10-14 |
| LMArena Creative Writing | 1489 Elo | 1430 Elo | source · 2026-10-15 |
| LMArena Instruction Following | 1484 Elo | 1469 Elo | source · 2026-10-15 |
| LMArena Long Query | 1492 Elo | 1464 Elo | source · 2026-10-15 |
| LMArena Math | 1510 Elo | 1470 Elo | source · 2026-10-15 |
| LMArena Text | 1488 Elo | 1484 Elo | source · 2026-10-15 |
| LMArena Text Factuality | 1475 Elo | 1465 Elo | source · 2026-10-15 |
| LMArena Vision | 1297 Elo | 1294 Elo | source · 2026-10-09 |
| LMArena Vision Diagrams | 1321 Elo | 1311 Elo | source · 2026-10-09 |
| LMArena Vision OCR | 1311 Elo | 1295 Elo | source · 2026-10-09 |
| LMArena WebDev | 1590 Elo | 1760 Elo | source · 2026-10-13 |
| LMArena WebDev Image-to-Web | 1603 Elo | 1703 Elo | source · 2026-10-09 |
| OTIS Mock AIME (Epoch AI) | 97.2% ± 1.9 | 100% ± 0 | source · 2026-08-14 |
| SimpleQA Verified (Epoch AI) | 69.2% ± 1.5 | 73.9% ± 1.4 | source · 2026-08-27 |
| Terminal-Bench (4.0) | 11.2% | 58.2% | source · 2026-10-16 |
Values are as published by each evaluator, in its own units; rows from different evaluators are not on a common scale.
task evidence
How they place when the question changes
Lower ranks are better. “Best” marks the strongest option in this shortlist for that task.