ShishirPatil/gorilla

Health v2 · maintenance only

Adoption not part of the score

13006 stars · 1400 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

Gorilla is a UC Berkeley research project for training and evaluating LLMs to make function/tool calls, including the OpenFunctions fine-tuned models and the Berkeley Function Calling Leaderboard (BFCL) benchmark. It also includes related work like RAFT for retriever-aware fine-tuning for domain-specific RAG.

Use cases

• evaluate how well LLMs call functions and tools
• fine-tune a model to generate API calls from natural language
• benchmark function-calling models on multi-turn agentic tasks
• compare LLMs on the Berkeley Function Calling Leaderboard
• improve RAG performance with retriever-aware fine-tuning
• train a model to select between multiple APIs

When to choose

• you need to benchmark or compare function-calling capabilities of LLMs
• you want to fine-tune an open model for reliable tool/API use
• you are building agentic systems that depend on accurate tool calls
• you need a research-grade evaluation dataset for function calling

When to avoid

• you just need a production LLM API client without evaluation or training
• you want a turnkey agent framework rather than research tooling
• your use case has nothing to do with LLM tool use or RAG

Facets

library · maturity active

llm-training benchmarking agent-framework rag machine-learning large-language-models artificial-intelligence apis developer-tools python cross-platform function-calling tool-use llm-evaluation berkeley-function-calling-leaderboard openfunctions fine-tuning api-benchmark ai-agents retrieval-augmented-generation

Member repositories

For agents

markdown · JSON · MCP: product_card(name="ShishirPatil/gorilla")