Claude 3.7 Sonnet (Thinking 16K): benchmark results
Provider: Anthropic. Released 2025-02-24. Access: API.
Aggregate Rating 1653 ± 28, rank #550 of 2287 rated models, from 14 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Generalization V1 (Lechmazur) | 1.73 | Avg Rank (lower is better) | 93.8 |
| PlatinumBench (MIT) | 0.93 | Category-balanced mean error rate (%) | 90.9 |
| Elimination Game (Lechmazur) | 5.28 | TrueSkill μ | 86.4 |
| Confabulation Leaderboard (Lechmazur) | 7.92 | Confabulation rate % (lower is better) | 80.8 |
| Arena-Hard Creative Writing | 63.9 | Win Rate (%) | 77.8 |
| Arena-Hard v2 | 59.8 | Win Rate (%) | 74.1 |
| MATH Level 5 | 86.25 | Accuracy (%) | 72.9 |
| Step Game (Lechmazur) | 2.76 | TrueSkill μ | 63.5 |
| NYT Connections Older Models | 19.7 | Score (%) | 59.9 |
| Arena-Hard v2 (GPT-4.1 Judge) | 47 | Win Rate (%) | 59.3 |
| Epoch AI - GPQA Diamond | 76.77 | Accuracy (%) | 51 |
| PGG-Bench (Lechmazur) | 9.96 | TrueSkill μ | 50 |
Interactive version: theaggregate.ai/model?slug=claude-3-7-sonnet-thinking-16k · How It Works · Data refreshed daily, snapshot 2026-11-22.