Claude Opus 4.8
AnthropicProprietaryReasoningVerifiedReleased May 28, 2026Input: text, image, file
Overall score
70.0
Rank 16 of 222 · 66.2–73.8
Context window
1M
1,000,000 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.
- 1Multimodal#591.2
- 2Knowledge#1687.5
- 3Math#184.4
- 4Instruction following#6575.4
- 5Coding#2075.1
- 6Agentic & tool use#2074.1
- 7Reasoning#2162.8
Benchmark results (36)
As reported in system cards, papers and public evaluations. Position compares this model with every other model reporting the same benchmark.
Agentic & tool use
| Terminal-Bench 2.1 | 74.6 | #36 of 68 |
| Gert Labs | 72.97 | #1 of 52 |
| BrowseComp | 84.3 | #15 of 50 |
| MCP Atlas | 82.2 | #9 of 42 |
| MazeBench | 0 | #20 of 42 |
| OSWorld-Verified | 83.4 | #5 of 35 |
| OSWorld 2.0 | 20.6 | #14 of 24 |
| Toolathlon | 59.9 | #2 of 22 |
| DeepSearchQA | 93.1 | #4 of 20 |
| ResearchClawBench | 21.1 | #1 of 19 |
| Terminal-Bench 3.0 | 21.1 | #5 of 13 |
Multimodal
| Design Arena Website | 1258 | #37 of 91 |
| CharXiv | 89.9 | #8 of 42 |
| ParseBench | 63.70 | #22 of 37 |
| ExtractBench (LlamaIndex)(lower is better) | 87.09 | #11 of 24 |
| ScreenSpot Pro | 87.9 | #2 of 15 |
| CharXiv w/o tools | 80.5 | #11 of 14 |
| OfficeQA Pro | 66.2 | #3 of 13 |
| SWE-bench Multimodal | 38.4 | #6 of 7 |
Coding
| SWE-bench Verified | 88.6 | #5 of 98 |
| SWE-bench Pro | 69.2 | #7 of 69 |
| FrontierCode 1.1 Main | 46.5 | #6 of 16 |
| PostTrainBench v1.1 | 32.9 | #8 of 15 |
Reasoning
| ARC-AGI-2 | 72.1 | #21 of 41 |
| ARC-AGI-1 | 92.50 | #18 of 36 |
| ARC-AGI-3 | 1.5 | #10 of 22 |
Knowledge
| GPQA | 93.6 | #5 of 117 |
| HLE | 57.9 | #6 of 69 |
| HLE w/o tools | 49.8 | #7 of 43 |
Math
| FrontierMath v2 (Tiers 1-3) | 80.000 | #15 of 64 |
| FrontierMath v2 (Tier 4) | 56.098 | #16 of 51 |
| USAMO 2026 | 96.7 | #2 of 3 |
Multilingual
| SWE Multilingual | 84.4 | #7 of 46 |
| INCLUDE | 87.6 | #2 of 10 |
Pricing by source
USD per 1M tokens.
| Source | Input | Output | Cached in |
|---|---|---|---|
| ✓ Official (Anthropic) checked Oct 17, 2026 · Claude Opus 4.8 | $5.00 | $25.00 | $0.50 |
| BenchLM list price | $5.00 | $25.00 | $0.50 |
| OpenRouter anthropic/claude-opus-4.8 | $5.00 | $25.00 | $0.50 |
Anthropic's May 28, 2026 Claude Opus 4.8 launch announcement says Opus 4.8 is available at the same price as Opus 4.7. BenchLM maps that to the current Opus API rate of $5 input / $25 output per million tokens.
Arena ratings
Human-preference ratings from LMArena. Best-rated variant shown.
| Text(high) | 1461 | #40 · 64.1K votes |
| WebDev (coding)(High) | 1556 | #41 · 18.5K votes |
| Vision(High) | 1295 | #23 · 18.3K votes |
| Search | 1204 | #19 · 106.3K votes |
| Documents(high) | 1475 | #12 · 11.6K votes |
Speed
Measured by us through OpenRouter (median of trials).
Not measured yet.
Related models
Same family
- Claude Opus 4.765.5
- Claude Opus 4.662.3
Similar score
- Kimi K370.0 · $6.00
- GPT-5.6 Terra70.2 · $4.50
- Qwen3.8 Max69.7 · —
- Step 5 Preview69.6 · $1.43
- GPT-5.569.2 · $11.25