Gemini 2.5 Pro
開発元 Google DeepMind · リリース日 2025-06-17
56.2
平均スコア
$1.25/1M
入力料金
$10.00/1M
出力料金
1.0M tokens (~524 books)
コンテキストウィンドウ
multimodal
タイプ
Tested on 42 benchmarks with 56.2% average. Top scores: Chatbot Arena Elo — Overall (1448.2%), Chatbot Arena Elo — Coding (1202.0%), MATH level 5 (95.6%).
ベンチマークスコア
| ベンチマーク | カテゴリ | スコア | Bar |
|---|---|---|---|
| Chatbot Arena Elo — Overall | arena | 1448.2 | |
| Chatbot Arena Elo — Coding | arena | 1202.0 | |
| MATH level 5 | math | 95.6 | |
| Fiction.LiveBench | knowledge | 91.7 | |
| OpenCompass — IFEval | language | 90.0 | |
| OpenCompass — AIME2025 | math | 88.7 | |
| HELM — MMLU-Pro | knowledge | 86.3 | |
| Lech Mazur Writing | knowledge | 86.0 | |
| OpenCompass — MMLU-Pro | knowledge | 85.8 | |
| HELM — WildBench | reasoning | 85.7 | |
| OTIS Mock AIME 2024-2025 | math | 84.7 | |
| OpenCompass — GPQA-Diamond | knowledge | 84.7 | |
| HELM — IFEval | language | 84.0 | |
| Aider polyglot | coding | 83.1 | |
| GeoBench | knowledge | 81.0 | |
| GPQA diamond | knowledge | 80.4 | |
| HELM — GPQA | knowledge | 74.9 | |
| OpenCompass — LiveCodeBenchV6 | coding | 71.3 | |
| CadEval | coding | 64.0 | |
| SWE-Bench verified | coding | 57.6 | |
| SimpleQA Verified | knowledge | 56.0 | |
| SimpleBench | reasoning | 54.9 | |
| WeirdML | coding | 54.0 | |
| DeepResearch Bench | knowledge | 49.7 | |
| AudioMultiChallenge | knowledge | 46.9 | |
| AudioMultiChallenge — Text Output | knowledge | 46.9 | |
| Balrog | knowledge | 43.3 | |
| HELM — Omni-MATH | math | 41.6 | |
| ARC-AGI | reasoning | 41.0 | |
| Artificial Analysis — Quality Index | speed | 34.6 | |
| Artificial Analysis — Agentic Index | speed | 32.7 | |
| Terminal Bench | coding | 32.6 | |
| Artificial Analysis — Coding Index | speed | 31.9 | |
| The Agent Company | agentic | 30.3 | |
| OpenCompass — HLE | knowledge | 21.1 | |
| Chess Puzzles | knowledge | 20.0 | |
| VPCT | knowledge | 19.6 | |
| HLE | knowledge | 17.7 | |
| FrontierMath-2025-02-28-Private | math | 14.1 | |
| ARC-AGI-2 | reasoning | 4.9 | |
| FrontierMath-Tier-4-2025-07-01-Private | math | 4.2 | |
| GSO-Bench | coding | 3.9 |