Benchmark
AgentBench
Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages.
All agentic benchmarksScores from Steel.dev Agent Leaderboards data.
- Models measured
- 3
- Scores recorded
- 5
- Sources
- 1
- IQ weight
- No weight published
Score distribution
- 57.2%
- Mean
- 58.2%
- Median
- 2.1%
- Std dev
- 53.2%
- 10th percentile
- 58.9%
- 90th percentile
- 53.2%
- Min
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4.5· Steel.dev Agent Leaderboards data | 58.9% |
| 2 | Claude Sonnet 4.5thinking· Steel.dev Agent Leaderboards data | 58.3% |
| 3 | Claude Sonnet 4thinking· Steel.dev Agent Leaderboards data | 58.2% |
| 4 | Claude Sonnet 4· Steel.dev Agent Leaderboards data | 57.4% |
| 5 | Claude Sonnet 3.7· Steel.dev Agent Leaderboards data | 53.2% |
Data credits: Steel.dev Agent Leaderboards, MIT (licence)