API-Bank
by Li et al. / Wuhan University · open-source · Last verified 2026-03-17
API-Bank is a comprehensive benchmark for evaluating tool-augmented LLMs across 73 APIs spanning daily-use categories. It tests three levels of difficulty — API retrieval, API calling, and plan-then-call — measuring both tool selection and execution correctness.
https://github.com/AlibabaResearch/DAMO-ConvAI/tree/main/api-bank ↗C+
C+—Average
Adoption: BQuality: AFreshness: B+Citations: BEngagement: F
Specifications
- License
- Apache-2.0
- Pricing
- open-source
- Capabilities
- evaluation, api-calling, tool-selection, agent-planning
- Integrations
- Use Cases
- model-evaluation, ai-agents, tool-augmented-llm
- API Available
- No
- Evaluated Models
- gpt-4o, claude-opus-4, llama-3-70b, toolllama
- Metrics
- accuracy, api-selection-accuracy, call-accuracy
- Methodology
- Three-level evaluation: Level 1 retrieves correct API from catalog; Level 2 generates correct API call; Level 3 plans and executes multi-step API sequences. Each level assessed with exact-match and functional correctness metrics.
- Last Run
- 2026-01-12
- Tags
- tool-use, api-call, agents, multi-step, planning
- Added
- 2026-03-17
- Completeness
- 100%
Index Score
58.8Adoption
62
Quality
85
Freshness
74
Citations
68
Engagement
0