Claude Sonnet 4.5: benchmark results
Anthropic Sonnet-tier Claude model for balanced reasoning, coding, and agentic tasks. Provider: Anthropic. Released 2025-09-29. Access: API.
Aggregate Rating 1732 ± 2, rank #142 of 1656 rated models, from 3956 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| 1D-Bench (Multi-Round) | 80.4 | Final score (0-100) after iterative component-level editing | 100 |
| ABC-Bench | 63.2 | Overall pass@1 (imported) | 100 |
| AGC-Bench - meta4xnli | 0.96 | Dataset z-score | 100 |
| AGC-Bench - slang_generation | 1.52 | Dataset z-score | 100 |
| AIDABench | 70.78 | Score (%) over all 603 tasks (category-weighted mean of the | 100 |
| AIDABench (Pass@1) - Data Visualization | 67.71 | Score (%) on the 116 data-visualization tasks, a single run | 100 |
| AIDABench (Pass@1) - File Generation | 49.43 | Score (%) on the 261 file-generation tasks, a single run (pa | 100 |
| AIDABench (Pass@1) - QA | 68.58 | Score (%) on the 226 question-answering tasks, a single run | 100 |
| AIDABench - File Generation | 66.28 | Score (%) on the 261 file-generation tasks, best of three ru | 100 |
| AIDABench - Hard | 50.54 | Score (%) on the 127 hard tasks (13 or more expert operation | 100 |
| AIDABench - Medium | 71.31 | Score (%) on the 298 medium tasks (7 to 12 expert operation | 100 |
| AIDABench - QA | 72.12 | Score (%) on the 226 question-answering tasks, best of three | 100 |
Interactive version: theaggregate.ai/model?slug=claude-sonnet-4-5 · How It Works · Data refreshed daily, snapshot 2026-10-20.