Claude Sonnet 5.5 (Max): benchmark results
Provider: Anthropic. Released 2026-09-28. Access: API.
Aggregate Rating 1946 ± 7, rank #21 of 2205 rated models, from 252 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Agent Arena - Bash Recovery | 15.04 | Bash Recovery (%) | 100 |
| Android Bench 2.0 - Long-horizon | 40 | Mean task pass rate (%) over 30 tasks | 100 |
| Compression-Bench - Nemotron-3.5 120B-A12B / 32 GiB (perplexity) | 1.8 | Perplexity of the compressed model on held-out BF16-generate | 100 |
| Compression-Bench - Qwen3.5 122B-A10B / 32 GiB (perplexity) | 2.88 | Perplexity of the compressed model on held-out BF16-generate | 100 |
| Compression-Bench - Qwen3.5 122B-A10B / 48 GiB (perplexity) | 2.79 | Perplexity of the compressed model on held-out BF16-generate | 100 |
| Compression-Bench - Qwen3.8 27B / 12 GiB (perplexity) | 3.25 | Perplexity of the compressed model on held-out BF16-generate | 100 |
| Compression-Bench - Qwen3.8 27B / 8 GiB (perplexity) | 3.36 | Perplexity of the compressed model on held-out BF16-generate | 100 |
| Epoch AI - Mirrorcode | 97.78 | Score | 100 |
| FrontierCS 2 - E-Graph Invariant Restoration | 67.31 | Task score (0–100) | 100 |
| FrontierCS 2 - Exact Sequence Alignment | 67.84 | Task score (0–100) | 100 |
| FrontierCS 2 - GPU DBSCAN Kernel Optimization | 73.38 | Task score (0–100) | 100 |
| FrontierCS 2 - GPU IVF-PQ Search Kernel Optimization | 80.99 | Task score (0–100) | 100 |
Interactive version: theaggregate.ai/model?slug=claude-sonnet-5-5-max · How It Works · Data refreshed daily, snapshot 2026-11-13.