TemporalBench: leaderboard
TemporalBench evaluates LLM-based agents on contextual and event-informed time-series tasks spanning multiple datasets and task types.
Metric: Overall MCQ Accuracy (%). Source: huggingface.co. 13 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | TimeClaw (deepseek-v4-pro) | 41.8 |
| 2 | Single LLM (deepseek-chat) | 36.47 |
| 3 | AgentScope (deepseek-chat) | 35.21 |
| 4 | CAMEL (deepseek-chat) | 35.11 |
| 5 | MetaGPT (deepseek-chat) | 34.42 |
| 6 | Single LLM (gpt-4o) | 34.11 |
| 7 | AgentScope (gpt-4o) | 32.91 |
| 8 | TimeCopilot (deepseek-chat) | 32.34 |
| 9 | CAMEL (gpt-4o) | 32.33 |
| 10 | MetaGPT (gpt-4o) | 31.56 |
Interactive version: theaggregate.ai/benchmark?slug=temporalbench · How It Works · Data refreshed daily, snapshot 2026-09-05.