Model
Claude Sonnet 4.5 (2025-09-29)
Anthropic · closed weights
Knowledge & Reasoning
Knowledge & Reasoning| Benchmark | Score | Metric | Harness | Source | Confidence | Link |
|---|---|---|---|---|---|---|
| Colombian Legal Reliability Benchmark | 80.3 | accuracy | N/A | Independent | medium | arxiv.org |
| CritPt | 1.1 | accuracy | N/A | Independent | low | epoch.ai |
| EnigmaEval | 6.0 | accuracy | N/A | Independent | medium | labs.scale.com |
| GameCommBench | 0.5 | analytical_completeness | N/A | Independent | medium | arxiv.org |
| GDPval | 42.5 | win_rate | N/A | Independent | low | epoch.ai |
| GPQA Diamond | 73.7 | accuracy | N/A | Independent | high | epoch.ai |
| HLE | 13.7 | accuracy | N/A | Independent | low | epoch.ai |
| SimpleQA Verified | 23.7 | accuracy | N/A | Independent | high | epoch.ai |
Coding
Coding| Benchmark | Score | Metric | Harness | Source | Confidence | Link |
|---|---|---|---|---|---|---|
| AptMQL-Bench | 67.9 | soft-ex | N/A | Independent | medium | arxiv.org |
| GSO | 14.7 | score_opt@1 | OpenHands | Independent | medium | epoch.ai |
| SciCode | 44.7 | score | N/A | Independent | medium | artificialanalysis.ai |
| SWE-bench Verified | 71.3 | resolved_rate | N/A | Independent | high | epoch.ai |
Agentic / Tool Use
Agentic / Tool Use| Benchmark | Score | Metric | Harness | Source | Confidence | Link |
|---|---|---|---|---|---|---|
| DeepResearch Bench | 47.5 | average_score | N/A | Independent | medium | evals.futuresearch.ai |
| OSWorld | 62.9 | score | claude-sonnet-4-5-20250929 (100 steps) | Independent | medium | os-world.github.io |
| OSWorld | 58.1 | score | claude-sonnet-4-5-20250929 (50 steps) | Independent | medium | os-world.github.io |
| OSWorld | 42.9 | score | claude-sonnet-4-5-20250929 (15 steps) | Independent | medium | os-world.github.io |
| Terminal-Bench | 46.5 | accuracy | CAMEL-AI | Independent | medium | tbench.ai |
| Terminal-Bench | 43.1 | accuracy | Goose | Independent | medium | tbench.ai |
| Terminal-Bench | 42.8 | accuracy | Terminus 2 | Independent | medium | tbench.ai |
| Terminal-Bench | 42.7 | accuracy | MAYA | Independent | low | epoch.ai |
| Terminal-Bench | 42.7 | accuracy | MAYA-V2 | Independent | medium | tbench.ai |
| Terminal-Bench | 42.6 | accuracy | OpenHands | Independent | medium | tbench.ai |
| Terminal-Bench | 42.5 | accuracy | Mini-SWE-Agent | Independent | medium | tbench.ai |
| Terminal-Bench | 40.1 | accuracy | Claude Code | Independent | medium | tbench.ai |
Multimodal
Multimodal| Benchmark | Score | Metric | Harness | Source | Confidence | Link |
|---|---|---|---|---|---|---|
| OmniVCBench | 49.4 | accuracy | N/A | Independent | medium | arxiv.org |
Safety & Alignment
Safety & Alignment| Benchmark | Score | Metric | Harness | Source | Confidence | Link |
|---|---|---|---|---|---|---|
| Cybench | 60.0 | unguided_%_solved | N/A | Self-reported | medium | assets.anthropic.com |
| Cybench | 55.0 | unguided_%_solved | N/A | Independent | medium | cybench.github.io |
| OMIT | 0.2 | obr | N/A | Independent | medium | arxiv.org |
Abstract Reasoning
Abstract ReasoningReproduced by the community
Loading community runs…