Claude Opus 4.6
Last updated: August 10, 2026
Specs
| Vendor | Anthropic |
|---|
| Released | September 29, 2025 |
|---|
| Context window | 500K tokens |
|---|
| Input | $5 / M tokens |
|---|
| Output | $25 / M tokens |
|---|
Benchmarks
| mmlu | 90.8 |
|---|
| gpqa | 62.1 |
|---|
| humanEval | 91.4 |
|---|
| arenaElo | 1402 |
|---|
| sweBench | 53.4 |
|---|
| sweBenchPro | 47 |
|---|
| liveCodeBench | 72 |
|---|
| aiderPolyglot | 74 |
|---|
| terminalBench | 43 |
|---|
| mmluPro | 84 |
|---|
| hle | 19 |
|---|
| aime | 84 |
|---|
| math500 | 95 |
|---|
| mmmu | 76 |
|---|
Strengths
- Battle-tested flagship before Opus 4.7
- Strong agentic coding and tool use
- 500K-token context on most deployments
Weaknesses
- Superseded by Opus 4.7 on reasoning and coding
- Smaller context than 4.7 (500K vs 1M)
- Same premium price as 4.7 without the upside
Best for
- Stable production pipelines already validated on 4.6
- Enterprise contracts pinned to 4.6 before the 4.7 upgrade
- Benchmark reproducibility on published 4.6 results
Head-to-head comparisons
Compare with other AI models