THUDM/AgentBench
A Comprehensive Benchmark to Evaluate LLMs as Agents (ICLR'24)
3.8k stars · 290 forks · Python
Healthy
Healthy: no known risks in the checks that ran.
Why
No rule fired.
What would change this
- A dormant maintenance status, a deprecated or archived package, a successor link, or a critical advisory would change this.
Maintenance
CompleteSteady
- Human commits, 30 days
- 0
- Human commits, 90 days
- 0
- Human commits, 365 days
- 8
- Bot share of commits, 365 days
- 0%
- Last commit
- Feb 8, 2026 (8 months ago)
- Last human commit
- Feb 8, 2026 (8 months ago)
- Authors, 90 days
- 0
- Last release
- none found
- Median PR merge time
- 31 hours
Responsiveness: of 3 recent issues, n/a got a maintainer reply. Median first response: n/a.
- Only 3 recent issues can be checked for a response, too few to measure response time.
Sources: Commits, Releases, Merged pull requests, Issues
People
Complete- Owner
- Organization
- Commits sampled
- 8 since Oct 15, 2025
- Authors for half of commits
- 1
- Authors for 80% of commits
- 2
- Contributors
- 12
Top committers
- Xiao9905: 63% of commits
- Moshe Kimhi: 38% of commits
Sources: Commits, Contributors
Stars
CompleteOrganic, confidence high
Anyone can buy stars for any repo; bursts alone don't prove the owner did.
- The star history looks organic.
Sources: How star history is analyzed
Risk
CompleteFlags
No risk flags.
License
- Now
- Apache-2.0 (OSI-approved)
- A year ago
- Apache-2.0 (OSI-approved)
Packages
No published packages found.
Sources: GitHub: THUDM/AgentBench
Badge
Add this to a README to show the current verdict:
[](https://reporadar.ai/THUDM/AgentBench)Earlier verdicts
- Oct 15, 2026: Healthy
- Oct 14, 2026: Healthy
Computed 3 hours ago · method 2026.10.2