Glossary
Berkeley Function Calling Leaderboard
A widely-cited benchmark for function-calling ability that checks emitted calls against expected ones by abstract-syntax-tree match and by executing them, across simple, parallel, and multi-turn cases.
The standard public yardstick for function callingagentsA pattern where a model emits a structured call (function name plus arguments), the runtime executes it, and the result returns as input on the model's next turn. Open full entry , maintained by the Gorilla project at UC Berkeley. It scores whether a model, given a set of tool schemas and a request, emits the right call with the right arguments. Two checking styles run side by side: an abstract-syntax-tree match that compares the emitted call structurally to a reference (right function, right required parameters, right types and values), and an executable check that actually runs the call and inspects the result.
The benchmarkevaluationA standardized dataset and scoring rubric used to compare model capability on a defined task, the unit of model evaluation since GLUE made the format the default. Open full entry separates cases that fail for different reasons. Simple single-call tasks test basic schema adherence; parallel and multiple-function tasks test whether a model picks correctly from a catalog and emits more than one call when needed; later additions test multi-turn interactions where the model gathers information across several steps. A relevance check also measures the opposite skill: not calling a tool when none of the offered tools fit, which a model eager to please tends to get wrong.
BFCL is one of the few tool-use benchmarks reported directly by model labs, which makes it a comparable axis across open checkpoints. On this site, tool-use scores appear on a model’s page only when the lab published the number with a date and source, the same benchmark discipline used for every other score.