GPT-5.1vsClaude Sonnet 4.5
Across 15 shared benchmarks, GPT-5.1 leads overall: GPT-5.1 wins 12, Claude Sonnet 4.5 wins 3, with 0 ties and an average score difference of +16.47.
GPT-5.1
OpenAI · 2025-11-12 · Reasoning model
Across 15 shared benchmarks, GPT-5.1 leads overall: GPT-5.1 wins 12, Claude Sonnet 4.5 wins 3, with 0 ties and an average score difference of +16.47.
OpenAI · 2025-11-12 · Reasoning model
Anthropic · 2025-09-30 · AI model
Grouped by capability, sorted by largest gap within each. 15 shared benchmarks.
| Benchmark | GPT-5.1 | Claude Sonnet 4.5 | Diff |
|---|---|---|---|
| ARC-AGI | 72.8025 / 65high | 25.5052 / 65 | +47.30 |
| HLE | 42.7038 / 149Thinking High (With Tools + Internet) | 7.10136 / 149 | +35.60 |
| GPQA Diamond | 88.1025 / 175Thinking High (No Tools) | 73.7094 / 175 | +14.40 |
| ARC-AGI-2 | 17.6032 / 58high | 3.8048 / 58 | +13.80 |
| Benchmark | GPT-5.1 | Claude Sonnet 4.5 | Diff |
|---|---|---|---|
| AIME2025 | 9428 / 106Thinking High (No Tools) | 3796 / 106 | +57 |
| FrontierMath | 26.7013 / 60Thinking High (With Tools) | 5.2038 / 60 | +21.50 |
| FrontierMath - Tier 4 | 12.5029 / 80Thinking High (With Tools) | 2.1056 / 80Normal (No Tools) | +10.40 |
| Simple Bench | 53.2010 / 27 |
| Benchmark | GPT-5.1 | Claude Sonnet 4.5 | Diff |
|---|---|---|---|
| Terminal Bench Hard | 432 / 13Thinking High (With Tools) | 338 / 13thinking + 使用工具 | +10 |
| τ²-Bench - Telecom | 95.6014 / 35Thinking High (With Tools) | 985 / 35thinking + 使用工具 | -2.40 |
| Benchmark | GPT-5.1 | Claude Sonnet 4.5 | Diff |
|---|---|---|---|
| SWE-Bench Pro - Public | 50.8024 / 36Thinking High (No Tools) | 43.6029 / 36thinking | +7.20 |
| SWE-bench Verified | 76.3025 / 103high | 823 / 103parallel_thinking + 使用工具 | -5.70 |
| Benchmark | GPT-5.1 | Claude Sonnet 4.5 | Diff |
|---|---|---|---|
| BrowseComp | 50.8034 / 43Thinking High (No Tools) | 24.1041 / 43thinking + 使用工具 | +26.70 |
| Benchmark | GPT-5.1 | Claude Sonnet 4.5 | Diff |
|---|---|---|---|
| Terminal Bench 2.0 | 47.6034 / 43Thinking High (With Tools) | 42.8038 / 43thinking + 使用工具 | +4.80 |
| Benchmark | GPT-5.1 | Claude Sonnet 4.5 | Diff |
|---|---|---|---|
| MMMU | 85.402 / 28Thinking High (No Tools) | 77.8014 / 28thinking | +7.60 |
| Field | GPT-5.1 | Claude Sonnet 4.5 |
|---|---|---|
| Publisher | OpenAI | Anthropic |
| Release date | 2025-11-12 | 2025-09-30 |
| Model type | Reasoning model | AI model |
| Architecture | Dense | Dense |
| Parameters | 0.0 | 0.0 |
| Context length | 400K | 1000K |
| Max output | 131072 | 65536 |
Prices use DataLearner records when available; missing fields are not inferred.
| Item | GPT-5.1 | Claude Sonnet 4.5 |
|---|---|---|
| Text input | 1.25 美元/100万 tokens | 3 美元/100 万tokens |
| Text output | 10 美元/100万 tokens | 15 美元/100 万tokens |
| Cache read | 0.125 美元/100万 tokens | 3.75 美元/100 万tokens |
| Cache write | Not public | 0.3 美元/100 万tokens |
On average across the 15 shared benchmarks, GPT-5.1 scores 16.47 higher.
Largest single-benchmark gap: AIME2025 — GPT-5.1 94 vs Claude Sonnet 4.5 37 (+57).
Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.
| 54.309 / 27 |
| -1.10 |
