Claude Sonnet 4.5vsGemini 2.5-Pro
Across 23 shared benchmarks, Gemini 2.5-Pro leads overall: Claude Sonnet 4.5 wins 9, Gemini 2.5-Pro wins 12, with 2 ties and an average score difference of -2.04.
Claude Sonnet 4.5
Anthropic · 2025-09-30 · AI model
Across 23 shared benchmarks, Gemini 2.5-Pro leads overall: Claude Sonnet 4.5 wins 9, Gemini 2.5-Pro wins 12, with 2 ties and an average score difference of -2.04.
Anthropic · 2025-09-30 · AI model
Google Deep Mind · 2025-06-05 · Reasoning model
Grouped by capability, sorted by largest gap within each. 23 shared benchmarks.
| Benchmark | Claude Sonnet 4.5 | Gemini 2.5-Pro | Diff |
|---|---|---|---|
| HLE | 7.10136 / 149 | 21.6089 / 149thinking | -14.50 |
| GPQA Diamond | 73.7094 / 175 | 86.4038 / 175thinking | -12.70 |
| ARC-AGI | 25.5052 / 65 | 3747 / 65thinking | -11.50 |
| MMLU Pro | 885 / 124thinking | 8618 / 124 | +2 |
| LiveBench | 70.5620 / 52 | 71.9213 / 52thinking | -1.36 |
| ARC-AGI-2 | 3.8048 / 58 | 4.9043 / 58thinking | -1.10 |
| Benchmark | Claude Sonnet 4.5 | Gemini 2.5-Pro | Diff |
|---|---|---|---|
| AIME2025 | 3796 / 106 | 8843 / 106thinking | -51 |
| IMO-ProofBench | 27.108 / 16thinking | 55.203 / 16thinking | -28.10 |
| IMO-ProofBench Advanced | 4.806 / 8thinking | 17.604 / 8thinking | -12.80 |
| Simple Bench | 54.309 / 27 |
| Benchmark | Claude Sonnet 4.5 | Gemini 2.5-Pro | Diff |
|---|---|---|---|
| τ²-Bench - Telecom | 985 / 35thinking + 使用工具 | 5432 / 35thinking + 使用工具 | +44 |
| Terminal Bench Hard | 338 / 13thinking + 使用工具 | 2512 / 13thinking + 使用工具 | +8 |
| Benchmark | Claude Sonnet 4.5 | Gemini 2.5-Pro | Diff |
|---|---|---|---|
| Terminal Bench 2.0 | 42.8038 / 43thinking + 使用工具 | 32.6043 / 43thinking + 使用工具 | +10.20 |
| Terminal-Bench | 2725 / 35normal + 使用工具 | 25.3028 / 35thinking | +1.70 |
| Benchmark | Claude Sonnet 4.5 | Gemini 2.5-Pro | Diff |
|---|---|---|---|
| LiveCodeBench | 5969 / 118 | 77.1032 / 118 | -18.10 |
| SWE-bench Verified | 823 / 103parallel_thinking + 使用工具 | 67.2063 / 103thinking | +14.80 |
| Benchmark | Claude Sonnet 4.5 | Gemini 2.5-Pro | Diff |
|---|---|---|---|
| BrowseComp | 24.1041 / 43thinking + 使用工具 | 7.8042 / 43thinking + 使用工具 | +16.30 |
| Benchmark | Claude Sonnet 4.5 | Gemini 2.5-Pro | Diff |
|---|---|---|---|
| IF Bench | 57.3019 / 27thinking + 使用工具 | 4926 / 27thinking + 使用工具 | +8.30 |
| Benchmark | Claude Sonnet 4.5 | Gemini 2.5-Pro | Diff |
|---|---|---|---|
| AA-LCR | 668 / 13thinking | 668 / 13thinking | — |
| Benchmark | Claude Sonnet 4.5 | Gemini 2.5-Pro | Diff |
|---|---|---|---|
| MMMU | 77.8014 / 28thinking | 829 / 28thinking | -4.20 |
| Benchmark | Claude Sonnet 4.5 | Gemini 2.5-Pro | Diff |
|---|---|---|---|
| GDPval-AA | 3915 / 20thinking | 2220 / 20thinking | +17 |
| Field | Claude Sonnet 4.5 | Gemini 2.5-Pro |
|---|---|---|
| Publisher | Anthropic | Google Deep Mind |
| Release date | 2025-09-30 | 2025-06-05 |
| Model type | AI model | Reasoning model |
| Architecture | Dense | Dense |
| Parameters | 0.0 | 0.0 |
| Context length | 1000K | 1000K |
| Max output | 65536 | 65536 |
Prices use DataLearner records when available; missing fields are not inferred.
| Item | Claude Sonnet 4.5 | Gemini 2.5-Pro |
|---|---|---|
| Text input | 3 美元/100 万tokens | 1.25 美元/100 万tokens |
| Text output | 15 美元/100 万tokens | 10 美元/100 万tokens |
| Cache read | 3.75 美元/100 万tokens | 0.125 美元/100 万tokens |
| Cache write | 0.3 美元/100 万tokens | Not public |
On average across the 23 shared benchmarks, Gemini 2.5-Pro scores 2.04 higher.
Largest single-benchmark gap: AIME2025 — Claude Sonnet 4.5 37 vs Gemini 2.5-Pro 88 (-51).
Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.
| 62.402 / 27thinking |
| -8.10 |
| FrontierMath | 5.2038 / 60 | 1123 / 60 | -5.80 |
| FrontierMath - Tier 4 | 2.1056 / 80Normal (No Tools) | 2.1056 / 80Normal (No Tools) | — |
