Claude Sonnet 4.5vsClaude Sonnet 4
Across 26 shared benchmarks, Claude Sonnet 4.5 leads overall: Claude Sonnet 4.5 wins 23, Claude Sonnet 4 wins 1, with 2 ties and an average score difference of +14.79.
Claude Sonnet 4.5
Anthropic · 2025-09-30 · Chat model
Claude Sonnet 4
Anthropic · 2025-05-23 · Reasoning model
Claude Sonnet 4.523 wins(88%)Ties2(4%)1 winClaude Sonnet 4
Benchmark scores
Grouped by capability, sorted by largest gap within each. 26 shared benchmarks.
General Knowledge
Claude Sonnet 4.5 5/6| Benchmark | Claude Sonnet 4.5 | Claude Sonnet 4 | Diff |
|---|---|---|---|
| HLE | 33.6080 / 172 | 9.60149 / 172 | +24 |
| ARC-AGI | 63.7035 / 68 | 4049 / 68 | +23.70 |
| ARC-AGI-2 | 13.6038 / 62 | 5.9046 / 62 | +7.70 |
| MMLU Pro | 887 / 132 | 8438 / 132 | +4 |
| LiveBench | 53.6983 / 115Normal (No Tools) | 50.9889 / 115Normal (No Tools) | +2.71 |
| GPQA Diamond | 83.4063 / 187 | 83.8062 / 187 | -0.40 |
Math and Reasoning
Claude Sonnet 4.5 3/5| Benchmark | Claude Sonnet 4.5 | Claude Sonnet 4 | Diff |
|---|---|---|---|
| AIME2025 | 1001 / 107 | 8551 / 107 | +15 |
| FrontierMath - Tier 4 | 2.1056 / 80Normal (No Tools) | 072 / 80Normal (No Tools) | +2.10 |
| FrontierMath | 5.2038 / 60 | 4.1041 / 60 | +1.10 |
| IMO-ProofBench | 27.108 / 16 | 27.108 / 16 | — |
| IMO-ProofBench Advanced | 4.806 / 8 | 4.806 / 8 | — |
Coding and Software Engineer
Claude Sonnet 4.5 4/4| Benchmark | Claude Sonnet 4.5 | Claude Sonnet 4 | Diff |
|---|---|---|---|
| CodeClash | 1,3891 / 8Normal (With Tools) | 1,2234 / 8Normal (With Tools) | +166 |
| LiveCodeBench | 7148 / 123 | 6659 / 123 | +5 |
| SWE-bench Verified | 828 / 112 | 80.2014 / 112 | +1.80 |
| SWE-Bench Pro - Public | 43.6047 / 54 | 42.7048 / 54 | +0.90 |
Agent Level Benchmark
Claude Sonnet 4.5 2/2| Benchmark | Claude Sonnet 4.5 | Claude Sonnet 4 | Diff |
|---|---|---|---|
| τ²-Bench - Telecom | 985 / 35 | 6529 / 35 | +33 |
| τ²-Bench | 84.709 / 43 | 5234 / 43 | +32.70 |
AI Agent - Tool Usage
Claude Sonnet 4.5 2/2| Benchmark | Claude Sonnet 4.5 | Claude Sonnet 4 | Diff |
|---|---|---|---|
| OSWorld-Verified | 61.4020 / 24 | 42.2022 / 24 | +19.20 |
| Terminal-Bench | 503 / 35 | 41.3010 / 35 | +8.70 |
Claw-style Agent Evaluation
Claude Sonnet 4.5 2/2| 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 |
Commonsense Reasoning
Claude Sonnet 4.5 1/1| Benchmark | Claude Sonnet 4.5 | Claude Sonnet 4 | Diff |
|---|---|---|---|
| Simple Bench | 54.3022 / 63Normal (No Tools) | 45.5034 / 63Thinking (No Tools) | +8.80 |
Instruction Following
Claude Sonnet 4.5 1/1| Benchmark | Claude Sonnet 4.5 | Claude Sonnet 4 | Diff |
|---|---|---|---|
| IF Bench | 57.3022 / 30 | 5523 / 30 | +2.30 |
Long Context
Claude Sonnet 4.5 1/1| Benchmark | Claude Sonnet 4.5 | Claude Sonnet 4 | Diff |
|---|---|---|---|
| AA-LCR | 6610 / 15 | 6512 / 15 | +1 |
Multimodal Understanding
Claude Sonnet 4.5 1/1| Benchmark | Claude Sonnet 4.5 | Claude Sonnet 4 | Diff |
|---|---|---|---|
| MMMU | 77.8015 / 29 | 76.5017 / 29 | +1.30 |
Productivity Knowledge
Claude Sonnet 4.5 1/1| Benchmark | Claude Sonnet 4.5 | Claude Sonnet 4 | Diff |
|---|---|---|---|
| GDPval-AA | 3916 / 21 | 3319 / 21 | +6 |
Specs
| Field | Claude Sonnet 4.5 | Claude Sonnet 4 |
|---|---|---|
| Publisher | Anthropic | Anthropic |
| Release date | 2025-09-30 | 2025-05-23 |
| Model type | Chat model | Reasoning model |
| Architecture | Dense | Dense |
| Parameters | Not available | Not available |
| Context length | 1000K | 200K |
| Max output | 64K | 64K |
API pricing
Prices use DataLearner records when available; missing fields are not inferred.
| Item | Claude Sonnet 4.5 | Claude Sonnet 4 |
|---|---|---|
| Text input | $3 / 1M tokens | $3 / 1M tokens |
| Text output | $15 / 1M tokens | $15 / 1M tokens |
| Cache read | $0.3 / 1M tokens | $0.3 / 1M tokens |
| Cache write | $3.75 / 1M tokens | $3.75 / 1M tokens |
Summary
- Claude Sonnet 4.5leads in:General Knowledge (5/6), Math and Reasoning (3/5), Coding and Software Engineer (4/4), Agent Level Benchmark (2/2), AI Agent - Tool Usage (2/2), Claw-style Agent Evaluation (2/2), Commonsense Reasoning (1/1), Instruction Following (1/1), Long Context (1/1), Multimodal Understanding (1/1), Productivity Knowledge (1/1)
On average across the 26 shared benchmarks, Claude Sonnet 4.5 scores 14.79 higher.
Largest single-benchmark gap: CodeClash — Claude Sonnet 4.5 1,389 vs Claude Sonnet 4 1,223 (+166).
Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.