Trainee-Bench - Success Rate: leaderboard
Share of Trainee-Bench tasks (The Agent's First Day, Fu et al., 2026) that the agent completes in full across the same 50 streaming workplace scenarios in the EvoEnv environment, where tasks carry deadlines and hidden clues must be found by exploration. Percent of tasks; higher is better.
Metric: Task Success Rate (%; tasks fully completed). Source: arxiv.org. Saturation forecast: Around February 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GLM-5 (Thinking) | 45 | #256 (GLM-5) |
| 2 | Gemini 3 Flash | 35 | #159 |
| 3 | MiniMax-M2.5 | 27 | #590 |
| 4 | Claude Sonnet 4 | 23 | #397 |
| 5 | GPT-5.1 | 23 | #225 |
| 6 | Grok 4 | 20 | #306 |
| 7 | Qwen 3 VL 235B A22B | 14 | #477 |
| 8 | GPT-4o | 13 | #741 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=trainee-bench-success-rate · How It Works · Data refreshed daily, snapshot 2026-11-12.