Claude Sonnet 4.6
Last updated: August 10, 2026
Specs
| Vendor | Anthropic |
|---|
| Released | October 1, 2025 |
|---|
| Context window | 200K tokens |
|---|
| Input | $3 / M tokens |
|---|
| Output | $15 / M tokens |
|---|
Benchmarks
| mmlu | 88.7 |
|---|
| gpqa | 59.4 |
|---|
| humanEval | 89 |
|---|
| arenaElo | 1358 |
|---|
| sweBench | 60 |
|---|
| sweBenchPro | 52 |
|---|
| liveCodeBench | 70 |
|---|
| aiderPolyglot | 72 |
|---|
| terminalBench | 42 |
|---|
| mmluPro | 80 |
|---|
| hle | 16 |
|---|
| aime | 80 |
|---|
| math500 | 93 |
|---|
| mmmu | 74 |
|---|
Strengths
- Best price-to-intelligence ratio in the Claude family
- Strong coding and tool use
- Fast enough for realtime UX
Weaknesses
- Context smaller than Opus 4.7 (200K vs 1M)
- Slightly behind on niche reasoning benchmarks
Best for
- Production chat agents
- Coding copilots
- Most MVP and SaaS workloads
Head-to-head comparisons
Compare with other AI models