Across 25 shared benchmarks, Claude Sonnet 4.5 leads overall: Claude Sonnet 4.5 wins 22, Claude Sonnet 4 wins 1, with 2 ties and an average score difference of +5.83.
Anthropic · 2025-09-30 · AI model
Anthropic · 2025-05-23 · Reasoning model
Grouped by capability, sorted by largest gap within each. 25 shared benchmarks.
| Benchmark | Claude Sonnet 4.5 | Claude Sonnet 4 | Diff |
|---|---|---|---|
| GPQA Diamond | 73.7094 / 175 | 68119 / 175 | +5.70 |
| LiveBench | 70.5620 / 52 | 65.4030 / 52 | +5.16 |
| MMLU Pro | 885 / 124thinking | 8435 / 124thinking | +4 |
| ARC-AGI-2 | 3.8048 / 58 | 1.3051 / 58 | +2.50 |
| ARC-AGI | 25.5052 / 65 | 23.8053 / 65 | +1.70 |
| HLE | 7.10136 / 149 | 5.52140 / 149 | +1.58 |
| Benchmark | Claude Sonnet 4.5 | Claude Sonnet 4 | Diff |
|---|---|---|---|
| Simple Bench | 54.309 / 27 | 45.5015 / 27thinking | +8.80 |
| FrontierMath - Tier 4 | 2.1056 / 80Normal (No Tools) | 072 / 80Normal (No Tools) | +2.10 |
| FrontierMath | 5.2038 / 60 | 4.1041 / 60 | +1.10 |
| AIME2025 | 3796 / 106 | 38 |
| Benchmark | Claude Sonnet 4.5 | Claude Sonnet 4 | Diff |
|---|---|---|---|
| LiveCodeBench | 5969 / 118 | 48.5092 / 118 | +10.50 |
| SWE-bench Verified | 823 / 103parallel_thinking + 使用工具 | 80.209 / 103parallel_thinking + 使用工具 | +1.80 |
| SWE-Bench Pro - Public | 43.6029 / 36thinking | 42.7030 / 36thinking | +0.90 |
| Benchmark | Claude Sonnet 4.5 | Claude Sonnet 4 | Diff |
|---|---|---|---|
| τ²-Bench - Telecom | 985 / 35thinking + 使用工具 | 6529 / 35thinking + 使用工具 | +33 |
| τ²-Bench | 7124 / 40normal + 使用工具 | 5233 / 40normal + 使用工具 | +19 |
| Benchmark | Claude Sonnet 4.5 | Claude Sonnet 4 | Diff |
|---|---|---|---|
| OSWorld-Verified | 61.4010 / 14thinking + 使用工具 | 42.2012 / 14thinking + 使用工具 | +19.20 |
| Terminal-Bench | 2725 / 35normal + 使用工具 | 2626 / 35normal + 使用工具 | +1 |
| Benchmark | Claude Sonnet 4.5 | Claude Sonnet 4 | Diff |
|---|---|---|---|
| Claw Bench | 88.1013 / 29Thinking (With Tools) | 77.8023 / 29Thinking (With Tools) | +10.30 |
| Pinch Bench | 88.204 / 37Thinking (With Tools) | 80.5022 / 37Thinking (With Tools) | +7.70 |
| Benchmark | Claude Sonnet 4.5 | Claude Sonnet 4 | Diff |
|---|---|---|---|
| IF Bench | 57.3019 / 27thinking + 使用工具 | 5520 / 27thinking + 使用工具 | +2.30 |
| Benchmark | Claude Sonnet 4.5 | Claude Sonnet 4 | Diff |
|---|---|---|---|
| AA-LCR | 668 / 13thinking | 6510 / 13thinking | +1 |
| Benchmark | Claude Sonnet 4.5 | Claude Sonnet 4 | Diff |
|---|---|---|---|
| MMMU | 77.8014 / 28thinking | 76.5016 / 28 | +1.30 |
| Benchmark | Claude Sonnet 4.5 | Claude Sonnet 4 | Diff |
|---|---|---|---|
| GDPval-AA | 3915 / 20thinking | 3318 / 20thinking | +6 |
| Field | Claude Sonnet 4.5 | Claude Sonnet 4 |
|---|---|---|
| Publisher | Anthropic | Anthropic |
| Release date | 2025-09-30 | 2025-05-23 |
| Model type | AI model | Reasoning model |
| Architecture | Dense | Dense |
| Parameters | 0.0 | 0.0 |
| Context length | 1000K | 200K |
| Max output | 65536 | 64000 |
Prices use DataLearner records when available; missing fields are not inferred.
| Item | Claude Sonnet 4.5 | Claude Sonnet 4 |
|---|---|---|
| Text input | 3 美元/100 万tokens | 3 美元/ 100万tokens |
| Text output | 15 美元/100 万tokens | 15 美元/100万tokens |
| Cache read | 3.75 美元/100 万tokens | 3.75 美元/ 100万tokens |
| Cache write | 0.3 美元/100 万tokens | 0.3 美元/ 100万tokens |
On average across the 25 shared benchmarks, Claude Sonnet 4.5 scores 5.83 higher.
Largest single-benchmark gap: τ²-Bench - Telecom — Claude Sonnet 4.5 98 vs Claude Sonnet 4 65 (+33).
Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.
| -1 |
| IMO-ProofBench | 27.108 / 16thinking | 27.108 / 16thinking | — |
| IMO-ProofBench Advanced | 4.806 / 8thinking | 4.806 / 8thinking | — |
