TemporalBench - T3: leaderboard
TemporalBench T3 family measuring MCQ accuracy on higher-level temporal reasoning over observed sequences and context.
Metric: Accuracy (%). Source: huggingface.co. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | TimeClaw (deepseek-v4-pro) | 42.18 | |
| 2 | TimeCopilot (deepseek-v4-pro) | 38.39 | |
| 3 | Single LLM (deepseek-chat) | 35.1 | |
| 4 | TimeCopilot (deepseek-chat) | 31.11 | |
| 5 | CAMEL (deepseek-chat) | 30.19 | |
| 6 | Single LLM (gpt-4o) | 29.24 | |
| 7 | AgentScope (gpt-4o) | 28.33 | |
| 8 | MetaGPT (deepseek-chat) | 28.05 | |
| 9 | TimeSeries Scientist (deepseek-chat) | 27.65 | |
| 10 | AgentScope (deepseek-chat) | 27.26 |
Interactive version: theaggregate.ai/benchmark?slug=temporalbench-t3 · How It Works · Data refreshed daily, snapshot 2026-10-20.