Claude Opus 4.5
AnthropicProprietaryNon-reasoningVerifiedReleased Nov 24, 2025Input: file, image, text
Overall score
54.0
Rank 78 of 222 · 41.4–66.7
Context window
200K
200,000 tokens
Price per 1M tokens
$5.00 in · $25.00 out
✓ Verified on Anthropic's pricing page
Blended price
$10.00
3:1 input:output tokens
Output speed
—
Not measured yet
Category scores
Scores per capability area with the model's rank among ranked models in that category.
- 1Multilingual#296.5
- 2Reasoning77.5
- 3Knowledge#8064.3
- 4Math#961.3
- 5Coding#6554.7
- 6Agentic & tool use52.0
- 7Multimodal#5048.6
- 8Instruction following#13131.9
Benchmark results (61)
As reported in system cards, papers and public evaluations. Position compares this model with every other model reporting the same benchmark.
Multimodal
| Design Arena Website | 1248 | #45 of 91 |
| MMMU-Pro | 70.6 | #40 of 48 |
| CharXiv | 68.5 | #40 of 42 |
| MMMU | 80.7 | #11 of 21 |
| MathVision | 74.3 | #20 of 20 |
| SimpleVQA | 65.7 | #11 of 17 |
| RealWorldQA | 77.0 | #13 of 16 |
| ERQA | 46.8 | #16 of 16 |
| ScreenSpot Pro | 45.7 | #15 of 15 |
| VideoMMMU | 84.4 | #9 of 12 |
| V* | 67.0 | #12 of 12 |
| OmniDocBench 1.5 | 87.7 | #9 of 11 |
| Video-MME (with subtitle) | 77.6 | #11 of 11 |
| MLVU (M-Avg) | 81.7 | #9 of 9 |
| CountBench | 90.6 | #6 of 8 |
| Video-MME (w/o subtitle) | 81.4 | #6 of 7 |
| AI2D_TEST | 87.7 | #7 of 7 |
| DynaMath | 79.7 | #6 of 6 |
| We-Math | 70.0 | #5 of 5 |
Agentic & tool use
| Gert Labs | 64.23 | #6 of 52 |
| MCP Atlas | 42.3 | #40 of 42 |
| Terminal-Bench 2.0 | 59.3 | #18 of 39 |
| Claw-Eval | 59.6 | #23 of 36 |
| OSWorld-Verified | 66.3 | #24 of 35 |
| JobBench | 32.3 | #23 of 34 |
| CyberGym | 50.6 | #23 of 25 |
| τ³-bench results | 70.2 | #7 of 23 |
| Toolathlon | 43.5 | #15 of 22 |
| WideResearch | 76.4 | #8 of 16 |
| VITA-Bench | 23.3 | #7 of 11 |
| QwenClawBench | 52.3 | #9 of 10 |
| DeepPlanning | 26.4 | #4 of 7 |
| MCP-Tasks | 71.8 | #3 of 5 |
| OSWorld | 66.3 | #1 of 2 |
Math
| FrontierMath v2 (Tiers 1-3) | 34.386 | #50 of 64 |
| FrontierMath v2 (Tier 4) | 4.878 | #46 of 51 |
| AIME26 | 95.1 | #14 of 29 |
| HMMT Feb 2026 | 85.3 | #14 of 24 |
| HMMT Feb 2025 | 92.9 | #9 of 13 |
| HMMT Nov 2025 | 93.3 | #5 of 12 |
| MMAnswerBench | 84.0 | #3 of 10 |
Multilingual
| SWE Multilingual | 77.5 | #20 of 46 |
| MMLU-ProX | 85.7 | #2 of 16 |
| INCLUDE | 86.2 | #3 of 10 |
| NOVA-63 | 56.7 | #5 of 7 |
| PolyMath | 79.0 | #3 of 6 |
| MAXIFE | 79.2 | #4 of 5 |
Knowledge
| GPQA | 87.0 | #40 of 117 |
| HLE | 30.8 | #41 of 69 |
| MMLU-Pro | 89.5 | #2 of 67 |
| SuperGPQA | 70.6 | #8 of 20 |
| MMLU-Redux | 96.6 | #1 of 15 |
| C-Eval | 92.2 | #4 of 10 |
Coding
| SWE-bench Verified | 80.9 | #12 of 98 |
| SWE-bench Pro | 57.1 | #38 of 69 |
| LiveCodeBench v6 | 84.8 | #18 of 46 |
| NL2Repo | 43.2 | #18 of 31 |
Reasoning
| LongBench v2 | 64.4 | #3 of 17 |
| AI-Needle | 74.0 | #1 of 4 |
Instruction following
| IFEval | 90.9 | #19 of 53 |
| IFBench | 58.0 | #33 of 42 |
Pricing by source
USD per 1M tokens.
| Source | Input | Output | Cached in |
|---|---|---|---|
| ✓ Official (Anthropic) checked Oct 17, 2026 · Claude Opus 4.5 | $5.00 | $25.00 | $0.50 |
| BenchLM list price | $5.00 | $25.00 | $0.50 |
| OpenRouter anthropic/claude-opus-4.5 | $5.00 | $25.00 | $0.50 |
Anthropic's Claude Opus 4.5 launch page lists Opus 4.5 at $5 input / $25 output per million tokens.
Arena ratings
Human-preference ratings from LMArena. Best-rated variant shown.
| WebDev (coding)(High, 32K, 20251101) | 1494 | #63 · 11K votes |
Speed
Measured by us through OpenRouter (median of trials).
Not measured yet.
Related models
Similar score
- Ternary Bonsai 2 27B54.0 · $0.18
- Claude Opus 4.5 Thinking54.1 · —
- Qwen3.6 Plus53.8 · $0.73
- Gemini 3 Flash54.3 · $1.13
- Qwen3.7 Plus53.8 · $0.56