Comparing 2 models
Claude Sonnet 5.5 vs Gemini 3.8 Flash
Benchmarks through
Verdict
Gemini 3.8 Flash places higher on 2 of the 3 tasks where both are ranked. Gemini 3.8 Flash costs 2.4x less per input token. Gemini 3.8 Flash accepts 1.0x more context.
- Choose Claude Sonnet 5.5 for Computer & Browser Use (#6).
- Choose Gemini 3.8 Flash for Charts & Diagrams (#1), Vision & Documents (#1).
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 | Claude Sonnet 5.5 | Gemini 3.8 Flash | Source |
|---|---|---|---|
| Design Arena Coding | 1366 Elo | 1306 Elo | source · 2026-10-16 |
| Design Arena UI | 1378 Elo | 1319 Elo | source · 2026-10-16 |
| Design Arena Website | 1330 Elo | 1303 Elo | source · 2026-10-16 |
| FrontierMath Tiers 1-3 (Epoch AI) | 88.8% ± 1.9 | 68.4% ± 2.8 | source · 2026-09-29 |
| GPQA Diamond (Epoch AI) | 95.6% ± 1.4 | 95.4% ± 1.4 | source · 2026-09-29 |
| LiveBench Agentic Coding | 56.3% | 54.2% | source · 2026-10-16 |
| LiveBench Coding | 91.4% | 72.5% | source · 2026-10-16 |
| LiveBench Data Analysis | 59.5% | 54.0% | source · 2026-10-16 |
| LiveBench Instruction Following | 56.8% | 81.4% | source · 2026-10-16 |
| LiveBench Language | 78.0% | 87.8% | source · 2026-10-16 |
| LiveBench Math | 96.1% | 91.6% | source · 2026-10-16 |
| LiveBench Reasoning | 91.6% | 89.3% | source · 2026-10-16 |
| LMArena Agent | 12.5 | 4 | source · 2026-10-14 |
| LMArena Agent Bash Recovery | 15 | -0.7 | source · 2026-10-08 |
| LMArena Agent Steerability | 12.3 | 2.1 | source · 2026-10-14 |
| LMArena Agent Task Outcome | 20 | 5.1 | source · 2026-10-14 |
| LMArena Agent Tool Hallucination | 0.5 | 0.5 | source · 2026-10-14 |
| LMArena Creative Writing | 1467 Elo | 1492 Elo | source · 2026-10-15 |
| LMArena Instruction Following | 1488 Elo | 1491 Elo | source · 2026-10-15 |
| LMArena Long Query | 1494 Elo | 1507 Elo | source · 2026-10-15 |
| LMArena Math | 1499 Elo | 1527 Elo | source · 2026-10-15 |
| LMArena Text | 1475 Elo | 1496 Elo | source · 2026-10-15 |
| LMArena Text Factuality | 1474 Elo | 1474 Elo | source · 2026-10-15 |
| LMArena Vision | 1266 Elo | 1289 Elo | source · 2026-10-09 |
| LMArena Vision Diagrams | 1320 Elo | 1333 Elo | source · 2026-10-09 |
| LMArena Vision OCR | 1280 Elo | 1317 Elo | source · 2026-10-09 |
| LMArena WebDev | 1774 Elo | 1585 Elo | source · 2026-10-13 |
| LMArena WebDev Image-to-Web | 1740 Elo | 1593 Elo | source · 2026-10-09 |
| OTIS Mock AIME (Epoch AI) | 100% ± 0 | 98.9% ± 0.9 | source · 2026-09-29 |
| SimpleQA Verified (Epoch AI) | 46.5% ± 1.6 | 69.7% ± 1.5 | source · 2026-09-29 |
| Terminal-Bench (4.0) | 61.8% | 19.1% | 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.