Benchmark leaderboard
The Berkeley Function Calling Leaderboard scores how accurately a model calls tools across simple, parallel, multi-turn and agentic function-calling tasks; the overall score is out of 100. This page lists every model ModelCap tracks with a published BFCL V4 result, its best evaluated configuration, the source's own rank, and the model's ModelCap Index position, API price and context window.
Snapshot as of 12 April 2026
Claude Opus 4.5 from Anthropic leads the BFCL V4 function-calling leaderboard among the 44 tracked models with 77.5 (Function calling configuration), per the source snapshot published 12 April 2026.
Sorted by published score, one row per model. Source: UC Berkeley (Gorilla).
BFCL V4 ranks models by function-calling accuracy. ModelCap ingests the board as published, matches each entry to a catalogue model with a reviewed identity, and shows the source's score, interval and rank unchanged. Where a model has an admitted result, it feeds the ModelCap Index as agent evidence alongside the other public boards; the methodology documents the weighting and the identity rules.
The Berkeley Function Calling Leaderboard scores how accurately a model calls tools across simple, parallel, multi-turn and agentic function-calling tasks; the overall score is out of 100. It is published by UC Berkeley (Gorilla).
Claude Opus 4.5 (Anthropic) holds the top BFCL V4 score among the models ModelCap tracks, at 77.5 as of the source snapshot published 12 April 2026.
44 tracked models have a published BFCL V4 result on ModelCap; the source board itself lists 83 entries. Each row shows the model's best evaluated configuration.
GLM 4.6 from Z.ai is the highest-scoring model with openly downloadable weights on this board, at 72.4.
Among the ten highest-scoring priced models, Qwen3 235B A22B Instruct 2507 has the lowest listed output price at $0.35 per 1M tokens while scoring 52.2.
The board ranks models by function-calling accuracy; higher scores are better. ModelCap shows the source's own score, interval and rank and never re-runs the evaluation.
Not on its own. The ModelCap Index combines several public capability sources with published uncertainty; BFCL V4 contributes as agent evidence where a model has an admitted result. The methodology page documents the weighting.
The newest BFCL V4 publication ModelCap holds is dated 12 April 2026. The page re-renders every minute from the sealed dataset, so it reflects the latest refresh of the source board.
Claude Haiku 4.5 is the highest-scoring model on this board that currently holds a ModelCap Index position (#70).