Up to four models
Claude Sonnet 5 vs DeepSeek-V4.1-Flash
On the index (GPT-4 = 100), Claude Sonnet 5 scores 164 and DeepSeek-V4.1-Flash 163. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
Claude Sonnet 5
164148–180
WeightsClosed
ReleasedJun 2026
Price in / out$2.0 / $10
Context1M
Core tests taken5 of 9
163147–179
WeightsOpen
ReleasedSep 2026
Price in / out$0.051 / $0.57
Context1.05M
Core tests taken6 of 9
Claude Sonnet 5 and DeepSeek-V4.1-Flash are too close to call (1 point apart, within the uncertainty), ahead on 3 of 5 shared core tests, at 22× the price.
Test by testthird-party results
Claude Sonnet 5
DeepSeek-V4.1-Flash
Every result as a table
| Test | Claude Sonnet 5 | DeepSeek-V4.1-Flash |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 90.5% | 89.8% |
| LiveBench Data Analysis | 71.7% | 79.3% |
| LiveBench IF | 63.9% | 70.0% |
| LiveBench Language | 75.0% | 81.2% |
| LiveBench Reasoning | 88.7% | 86.7% |
| SimpleBench | 60.6% | 66.7% |
| SimpleQA Verified | 33.7% | 42.9% |
| Maths | ||
| FrontierMath (tiers 1-3) | 65.6% | 67.4% |
| AIME-style maths (OTIS mock) | 94.7% | 98.3% |
| LiveBench Mathematics | 92.9% | 93.3% |
| ProofBench v1.1 | 77.0% | 54.0% |
| Chess puzzles | 45.0% | 48.0% |
| FrontierMath (tier 4) | 29.3% | 26.8% |
| Surface Evolver Bench | 60.0% | 46.3% |
| Coding | ||
| Terminal-Bench 4.0 | 10.6% | 19.7% |
| LiveBench Agentic Coding | 59.4% | 77.3% |
| LiveBench Coding | 80.7% | 80.0% |
| Code Migration | 44.4% | 45.6% |
| CyberBench v1.1 | 70.3% | 73.7% |
| IOI | 45.0% | 40.3% |
| ProgramBench | 0.0% | 0.5% |
| Vibe Code Bench 1-100 | 13.8% | 16.4% |
| Vibe Code Bench v1.1 | 81.3% | 84.7% |
| MirrorCode | 63.3% | 64.4% |
| Agents | ||
| APEX-Agents | 54.5% | 39.5% |
| SkillsBench | 46.5% | 69.8% |
| SRE Bench | 0.4% | 0.8% |
| Terminal-Bench-Science | 5.7% | 0.0% |
| DTBench | 92.5% | 89.9% |
| LMCA | 50.0% | 47.0% |
| Mystery game puzzles | 35.0% | 43.0% |
| Professional work | ||
| Finance Agent (v2) | 53.9% | 53.5% |
| EMB | 66.3% | 57.2% |
| LegalBench | 83.9% | 83.3% |
| Legal Research Bench | 41.8% | 41.3% |
| MedCode | 47.5% | 41.2% |
| MedScribe | 76.0% | 85.5% |
| Public Benefits Bench v1.1 | 66.0% | 64.3% |
| Tax Agent Bench | 29.4% | 26.6% |
Tests at least two of these models have taken. Bold rows are the core tests.