Up to four models
Claude Sonnet 5 vs Claude Sonnet 4.6
On the index (GPT-4 = 100), Claude Sonnet 5 scores 164 and Claude Sonnet 4.6 159. 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
Claude Sonnet 4.6
159143–175
WeightsClosed
ReleasedFeb 2026
Price in / out$3.0 / $15
Context1M
Core tests taken5 of 9
Claude Sonnet 5 leads Claude Sonnet 4.6 by 5 points, ahead on 3 of 3 shared core tests, and costs 1.5× less.
Test by testthird-party results
Claude Sonnet 5Claude Sonnet 4.6
Every result as a table
| Test | Claude Sonnet 5 | Claude Sonnet 4.6 |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 90.5% | 87.4% |
| LiveBench Data Analysis | 71.7% | 78.0% |
| LiveBench IF | 63.9% | 63.2% |
| LiveBench Language | 75.0% | 76.1% |
| LiveBench Reasoning | 88.7% | 84.8% |
| GPQA Diamond (Vals AI) | 88.9% | 85.6% |
| MMLU Pro | 87.5% | 87.3% |
| MMMU Pro | 83.0% | 83.6% |
| GeoBench | 74.0% | 73.0% |
| SimpleQA Verified | 33.7% | 35.5% |
| Maths | ||
| AIME-style maths (OTIS mock) | 94.7% | 85.8% |
| LiveBench Mathematics | 92.9% | 87.0% |
| Chess puzzles | 45.0% | 23.0% |
| Coding | ||
| LiveBench Agentic Coding | 59.4% | 42.6% |
| LiveBench Coding | 80.7% | 79.3% |
| Code Migration | 44.4% | 39.9% |
| LiveCodeBench | 82.4% | 82.1% |
| ProgramBench | 0.0% | 0.5% |
| SWE-bench | 79.6% | 77.4% |
| Vibe Code Bench v1.1 | 81.3% | 51.5% |
| DeepSWE | 53.8% | 29.9% |
| WeirdML | 69.0% | 66.1% |
| Agents | ||
| APEX-Agents | 54.5% | 43.0% |
| SkillsBench | 46.5% | 49.0% |
| DTBench | 92.5% | 89.9% |
| LMCA | 50.0% | 46.5% |
| Mystery game puzzles | 35.0% | 16.0% |
| Professional work | ||
| Finance Agent (v2) | 53.9% | 51.0% |
| CorpFin v2 | 67.0% | 65.3% |
| EMB | 66.3% | 60.2% |
| LegalBench | 83.9% | 82.1% |
| Legal Research Bench | 41.8% | 38.5% |
| MortgageTax | 70.0% | 67.7% |
| Public Benefits Bench v1.1 | 66.0% | 62.5% |
| Tax Agent Bench | 29.4% | 25.8% |
| TaxEval v2 | 75.6% | 77.1% |
Tests at least two of these models have taken. Bold rows are the core tests.