Claude Sonnet 4
| Released | 2025-05-22 |
|---|---|
| Price per 1M | $3.00 in · $15.00 out |
| Overall score | 47.1 · rank 134 |
| Context window | 200,000 tokens |
| Provider | Anthropic |
| Weights | Proprietary |
Scores
| Dimension | Score | Rank | Benchmarks |
|---|---|---|---|
| Overall: Balanced | 47.1 | 134 | 27 |
| Overall: Coding | 47.1 | 131 | 14 |
| Overall: Reasoning | 46.1 | 149 | 13 |
| Overall: Agentic work | 48.4 | 126 | 14 |
| Overall: Frontend | 46.3 | 124 | 8 |
| Coding | 46.0 | 148 | 4 |
| Agents | 49.3 | 110 | 3 |
| Reasoning | 48.9 | 123 | 7 |
| Math | 44.4 | 150 | 4 |
| Knowledge | 42.6 | 129 | 2 |
| General | 50.3 | 103 | 3 |
| Multimodal | 47.3 | 85 | 4 |
| Value vs price | −5.9 | 163 |
Configurations
| Configuration | Balanced | Benchmarks |
|---|---|---|
| Unspecified setting | 46.1 ±1.1 | 25 |
| Thinking | 48.0 ±1.2 | 16 |
Evidence
Coding 46.0 ±2.0
| Benchmark | Result | Source | Date |
|---|---|---|---|
| Aider Polyglot | 61.3% | Epoch AI | 2026-11-11 |
| GSO | 4.9% | Epoch AI | 2026-11-11 |
| SciCode | 40.1% | Epoch AI | 2026-11-11 |
| LiveCodeBench (Vals run) | 62.4% | Vals AI | 2026-09-01 |
Agents 49.3 ±3.4
| Benchmark | Result | Source | Date |
|---|---|---|---|
| METR task suite | 62.0% | Epoch AI | 2026-11-11 |
| DeepResearch Bench | 46.6% | Epoch AI | 2026-11-11 |
| OSWorld-Verified | 43.9% | OSWorld-Verified | 2025-07-28 |
| OSWorld (v1) | 43.9% | Epoch AI | 2026-11-11 |
Reasoning 48.9 ±2.0
| Benchmark | Result | Source | Date |
|---|---|---|---|
| ARC-AGI-2 | 5.9% | Epoch AI | 2026-11-11 |
| ARC-AGI-2 (ARC Prize verified) | 5.9% | ARC Prize | 2026-11-11 |
| CritPt | 0.3% | Epoch AI | 2026-11-11 |
| ARC-AGI-1 (ARC Prize verified) | 40.0% | ARC Prize | 2026-11-11 |
| ARC-AGI-1 | 40.0% | Epoch AI | 2026-11-11 |
| EnigmaEval | 3.1% | Epoch AI | 2026-11-11 |
| GPQA Diamond | 79.2% | Epoch AI | 2025-05-26 |
| GPQA Diamond (Vals run) | 75.0% | Vals AI | 2026-09-01 |
| Kagi Reasoning Benchmark | 73.0% | Kagi | 2026-11-11 |
| SimpleBench | 45.5% | SimpleBench | 2026-11-11 |
| SimpleBench | 45.5% | Epoch AI | 2026-11-11 |
Math 44.4 ±2.1
| Benchmark | Result | Source | Date |
|---|---|---|---|
| MATH Level 5 | 84.4% | Epoch AI | 2025-05-22 |
| FrontierMath Tiers 1-3 (2025 set) | 4.1% | Epoch AI | 2025-07-04 |
| FrontierMath Tier 4 (2025 set) | 0.0% | Epoch AI | 2025-07-01 |
| OTIS Mock AIME 2024-2025 | 71.1% | Epoch AI | 2025-05-22 |
Knowledge 42.6 ±1.3
| Benchmark | Result | Source | Date |
|---|---|---|---|
| Humanity's Last Exam | 7.8% | Epoch AI | 2026-11-11 |
| MMLU-Pro (Vals run) | 83.9% | Vals AI | 2026-09-01 |
General 50.3 ±3.2
| Benchmark | Result | Source | Date |
|---|---|---|---|
| LMArena Text | 1,402 | LMArena | 2026-11-09 |
| EQ-Bench Creative Writing v3 | 1,483 | EQ-Bench | 2026-11-11 |
| EQ-Bench Longform Creative Writing | 58.4% | EQ-Bench | 2026-11-11 |
Multimodal 47.3 ±3.8
| Benchmark | Result | Source | Date |
|---|---|---|---|
| LMArena Vision | 1,206 | LMArena | 2026-11-09 |
| MMMU-Pro (Vals run) | 74.9% | Vals AI | 2026-09-01 |
| VPCT | 34.0% | Epoch AI | 2026-11-11 |
| GeoBench | 37.0% | Epoch AI | 2026-11-11 |
Results not used in the score
| Benchmark | Result | Source | Why not used |
|---|---|---|---|
| ARC-AGI-1 | 23.8% | Epoch AI | A stronger configuration result was used |
| ARC-AGI-1 (ARC Prize verified) | 23.8% | ARC Prize | A stronger configuration result was used |
| ARC-AGI-2 | 1.3% | Epoch AI | A stronger configuration result was used |
| ARC-AGI-2 (ARC Prize verified) | 1.3% | ARC Prize | A stronger configuration result was used |
| Aider Polyglot | 56.4% | Epoch AI | A stronger configuration result was used |
| GPQA Diamond | 66.7% | Epoch AI | A stronger configuration result was used |
| GPQA Diamond (Vals run) | 69.4% | Vals AI | A stronger configuration result was used |
| GeoBench | 32.0% | Epoch AI | A stronger configuration result was used |
| Kagi Reasoning Benchmark | 55.9% | Kagi | A stronger configuration result was used |
| LMArena Text | 1,391 | LMArena | A stronger configuration result was used |
| LMArena Vision | 1,189 | LMArena | A stronger configuration result was used |
| LiveCodeBench (Vals run) | 59.7% | Vals AI | A stronger configuration result was used |
| MMLU-Pro (Vals run) | 79.4% | Vals AI | A stronger configuration result was used |
| MMMU-Pro (Vals run) | 72.4% | Vals AI | A stronger configuration result was used |
| OSWorld (v1) | 41.4% | Epoch AI | A stronger configuration result was used |
| OSWorld (v1) | 31.2% | Epoch AI | A stronger configuration result was used |
| OSWorld-Verified | 31.2% | OSWorld-Verified | A stronger configuration result was used |
| OSWorld-Verified | 41.4% | OSWorld-Verified | A stronger configuration result was used |
| OTIS Mock AIME 2024-2025 | 28.9% | Epoch AI | A stronger configuration result was used |
| SimpleBench | 45.5% | Epoch AI | A stronger configuration result was used |
| VPCT | 30.0% | Epoch AI | A stronger configuration result was used |
Names matched to this model
claude-sonnet-4, claude-sonnet-4-20250514, claude-strudel-v8-p