Anthropic
availableShows if the model has enough results for an index.Claude Opus 4.5
Claude Opus 4.5 is a non-reasoning model from Anthropic. 53 benchmarks count toward its score, in 7 categories.
IndexOverall score out of 100.52.1 ±2.7
CoverageShare of the index weight with results.95%
SpeedOutput tokens per second.40/s
Input / 1MUS dollars per 1M input tokens.$5 batch $2.5
Output / 1MUS dollars per 1M output tokens.$25 batch $12.5 US dollars per 1M output tokens in a batch.
ContextMaximum tokens in one request.200K
EloLMArena rating and rank.1450 (#44)
The index is a score out of 100. The ± range shows how much it can change. Batch work costs less.
70,013 votes. Elo shows what people prefer. It does not change the score.
CapabilitiesScore per category, out of 100.
Out of 100Results
53 counted| BenchmarkThe test name. | CategoryThe capability that the test measures. | ResultThe score from the publisher. | IndexThis result as a score out of 100. | RunThe settings of the run. | DateDate of the result. | Published byThe source of the result. |
|---|---|---|---|---|---|---|
| MMLU-Redux | Knowledge | 96.6% | 56.5 | — | — | Qwen |
| AIME 2026 | Math | 95.1% | 54.3 | — | — | Qwen |
| MGSM | Multilingual | 94.8% | — | — | 9 Jan 2026 | Vals AI |
| Harvard-MIT Mathematics Tournament November 2025 | Math | 93.3% | — | — | — | Qwen |
| Harvard-MIT Mathematics Tournament February 2025 | Math | 92.9% | 51.6 | — | — | Qwen |
| τ²-bench Telecom | Agentic | 92.3% | 65.1 | high effort · Sierra | 2 Mar 2026 | Sierra Research |
| C-Eval | Knowledge | 92.2% | — | — | — | C-Eval authors |
| Instruction-Following Eval | Instruction | 90.9% | 47.1 | — | — | Jeffrey Zhou et al. |
| Massive Multitask Language Understanding Professional | Knowledge | 89.5% | 61.6 | — | — | Yubo Wang et al. |
| Artificial Analysis MMLU-Pro | Knowledge | 88.9% | — | — | — | Artificial Analysis |
| Graduate-Level Google-Proof Q&A | Knowledge | 87.0% | 58.6 | — | — | David Rein et al. |
| τ²-Bench Tool-Agent-User Evaluation | Agentic | 86.3% | 60.7 | — | — | Victor Barres et al. |
| MMLU-ProX | Multilingual | 85.7% | — | — | — | MMLU-ProX authors |
| MMLU Pro | Knowledge | 85.6% | 55.4 | — | 1 Sept 2026 | Vals AI |
| GPQA diamond | Knowledge | 85.5% | 57.2 | — | — | Epoch AI |
| Harvard-MIT Mathematics Tournament February 2026 | Math | 85.3% | 54.0 | — | — | Qwen |
| LiveCodeBench v6 | Coding | 84.8% | 54.2 | — | — | LiveCodeBench maintainers |
| VideoMMMU | Multimodal | 84.4% | — | — | — | Qwen |
| MMAnswerBench | Math | 84.0% | — | — | — | Qwen |
| τ²-bench Airline | Agentic | 84.0% | 59.1 | high effort · Sierra | 2 Mar 2026 | Sierra Research |
| OTIS Mock AIME 2024-2025 | Math | 81.7% | 57.2 | — | — | Epoch AI |
| MMMU Pro | Multimodal | 81.1% | 59.3 | — | 1 Sept 2026 | Vals AI |
| Artificial Analysis GPQA Diamond | Knowledge | 81.0% | 51.6 | — | — | Artificial Analysis |
| Software Engineering Benchmark Verified | Coding | 80.9% | 62.4 | — | — | Carlos E. Jimenez et al. |
| τ²-bench Retail | Agentic | 79.6% | 55.9 | high effort · Sierra | 30 Apr 2026 | Sierra Research |
| GPQA Diamond | Knowledge | 79.5% | 51.7 | — | 1 Sept 2026 | Vals AI |
| AIME | Math | 76.9% | 49.6 | — | 16 Apr 2026 | Vals AI |
| SWE-bench Verified | Coding | 76.8% | 59.1 | medium effort · live-SWE-agent | 1 Sept 2026 | SWE-bench team |
| SWE-Bench verified | Coding | 76.7% | 59.0 | — | — | Epoch AI |
| WideResearch | Agentic | 76.4% | 62.0 | — | — | Qwen |
| LiveCodeBench | Coding | 75.0% | 52.5 | — | 1 Sept 2026 | Vals AI |
| MathVision | Multimodal | 74.3% | — | — | — | Qwen |
| AI-Needle | Reasoning | 74.0% | — | — | — | Qwen |
| MCP-Tasks | Agentic | 71.8% | — | — | — | Qwen |
| Artificial Analysis MMMU-Pro | Multimodal | 71.2% | 54.1 | — | — | Artificial Analysis |
| Artificial Analysis Long Context Reasoning | Reasoning | 70.7% | 57.2 | — | — | Artificial Analysis |
| SWE-bench Multilingual | Coding | 70.7% | — | mini-SWE-agent | 2 Sept 2026 | SWE-bench team |
| Massive Multi-discipline Multimodal Understanding Pro | Multimodal | 70.6% | 42.2 | — | — | MMMU-Pro authors |
| SuperGPQA: Scaling LLM Evaluation Across 285 Graduate Disciplines | Knowledge | 70.6% | 55.9 | — | — | Xiaoxuan Du et al. |
| τ³-Bench Tool-Agent-User Evaluation | Agentic | 70.2% | 54.8 | — | — | Sierra Research |
| CharXiv Reasoning | Multimodal | 68.5% | 44.1 | — | — | CharXiv authors |
| V* | Multimodal | 67.0% | 21.0 | — | — | Z.AI |
| OSWorld-Verified | Agentic | 66.3% | 55.4 | — | — | Tianbao Xie et al. |
| OSWorld | Agentic | 66.3% | — | — | — | Z.AI |
| LongBench v2 | Reasoning | 64.4% | — | — | — | LongBench v2 authors |
| Gert Labs Composite Game Benchmark | Agentic | 64.2% | 69.4 | — | — | Gert Labs |
| Claw-Eval | Agentic | 59.6% | 52.1 | — | — | Bowen Ye et al. |
| Terminal-Bench 2.0 | Agentic | 58.4% | 64.4 | — | 4 Jun 2026 | Vals AI |
| Instruction Following Benchmark | Instruction | 58.0% | 33.6 | — | — | Benchmark authors |
| SWE-bench Pro | Coding | 57.1% | 59.2 | — | — | Xiang Deng et al. |
| NOVA-63 | Multilingual | 56.7% | — | — | — | Qwen |
| Terminal-Bench 1.0 | Agentic | 56.3% | 59.2 | — | 12 Jan 2026 | Vals AI |
| SWE-bench (full test split) | Coding | 52.6% | — | Sonar Foundation Agent | 19 Dec 2025 | SWE-bench team |
| QwenClawBench | Agentic | 52.3% | 51.7 | — | — | Qwen |
| CyberGym | Agentic | 50.6% | 50.8 | — | — | Zhun Wang et al. |
| ScreenSpot Pro | Multimodal | 45.7% | 26.2 | — | — | Kaixin Li et al. |
| SimpleQA Verified | Knowledge | 45.7% | 63.9 | — | — | Epoch AI |
| Toolathlon | Agentic | 43.5% | 58.5 | — | — | OpenAI |
| NL2Repo | Coding | 43.2% | 58.8 | — | — | MiniMax |
| Artificial Analysis IFBench | Instruction | 43.0% | 32.7 | — | — | Artificial Analysis |
| MCP Atlas | Agentic | 42.3% | 40.1 | — | — | OpenAI |
| Artificial Analysis Omniscience Accuracy | Knowledge | 40.9% | 64.5 | — | — | Artificial Analysis |
| FrontierMath-Tiers-1-3-v2-Private | Math | 34.4% | 50.8 | — | — | Epoch AI |
| JobBench | Agentic | 32.3% | 56.3 | — | — | Yuetai Li et al. |
| Humanity's Last Exam | Knowledge | 30.8% | 54.9 | — | — | Center for AI Safety et al. |
| Furniture Assembly | Reasoning | 28.3% | 59.7 | — | — | Epoch AI |
| DeepPlanning | Agentic | 26.4% | — | — | — | DeepPlanning authors |
| τ²-bench Banking | Agentic | 24.7% | 16.6 | high effort · Sierra | 4 Aug 2026 | Sierra Research |
| Artificial Analysis Intelligence Index | Knowledge | 23.7% | 52.0 | — | — | Artificial Analysis |
| IOI v1 | Coding | 23.6% | 53.3 | — | 9 Aug 2026 | Vals AI |
| VITA-Bench | Agentic | 23.3% | 42.2 | — | — | Meituan LongCat Team |
| Mystery Game Puzzles | Reasoning | 22.0% | 57.3 | — | — | Epoch AI |
| FrontierMath-2025-02-28-Private | Math | 20.7% | 49.9 | — | — | Epoch AI |
| EBR-bench | Reasoning | 14.3% | 59.1 | — | — | Epoch AI |
| Artificial Analysis Humanity's Last Exam | Knowledge | 13.2% | 38.9 | — | — | Artificial Analysis |
| Chess Puzzles | Reasoning | 8.0% | 33.9 | — | — | Epoch AI |
| FrontierMath-Tier-4-v2-Private | Math | 4.9% | 50.3 | — | — | Epoch AI |
| FrontierMath-Tier-4-2025-07-01-Private | Math | 4.2% | 48.0 | — | — | Epoch AI |
| Critical Physics Tasks | Reasoning | 0.3% | 39.0 | — | — | Artificial Analysis |
53 benchmarks count, from 63 of 79 results. A grey row does not count. Too few models took that benchmark.