gorilla

Why we included this project

The Gorilla project tackles a specific problem that keeps surfacing in real agent work: getting an LLM to pick the right tool and call it correctly. It releases the OpenFunctions fine-tuned model checkpoints, and these days it's best known for maintaining the Berkeley Function Calling Leaderboard (BFCL), a benchmark that compares how well different models handle tool calls in realistic agentic settings. If you're building agent workflows where a model has to call functions, do web searches, manage memory, or recover from errors, the repo gives you reference models and a concrete way to measure the candidates you're considering. It's research-led, so expect datasets, weights, and evaluation code rather than a turnkey service. Teams weighing function-calling quality before settling on a stack will find the leaderboard and its changelog a practical reference.