ShishirPatil/gorilla
A research project training and evaluating LLMs to make API function calls and use tools effectively.
Velocity · 7d
+12
★ / day
Trend
→steady
star history
Gorilla provides training methodologies and evaluation benchmarks for LLMs on function calling tasks. The Berkeley Function Calling Leaderboard (BFCL) serves as a comprehensive benchmark testing LLMs on API interactions, multi-hop reasoning, web search, and memory management in agentic settings. The project supports multiple API providers including OpenAI and Claude.