BFCL-v3

Multilingual: no

Berkeley Function Calling Leaderboard v3 (BFCL-v3) is an advanced benchmark that evaluates large language models' function calling capabilities through multi-turn and multi-step interactions. It introduces extended conversational exchanges where models must retain contextual information across turns and execute multiple internal function calls for complex user requests. The benchmark includes 1000 test cases across domains like vehicle control, trading bots, travel booking, and file system management, using state-based evaluation to verify both system state changes and execution path correctness.

Methodology

Imported from llm-stats public benchmark metadata. Modality: text. Max score: 1. Categories: agents, finance, general, reasoning, structured_output, tool_calling. Language: en. Verified by llm-stats: no.

Leaderboard

Showing 18 of 18 results

GLM-4.5 self-reported llm-stats

77.8%
GLM-4.5-Air self-reported llm-stats

76.4%
LongCat-Flash-Thinking self-reported llm-stats

74.4%
Qwen3-Next-80B-A3B-Thinking self-reported llm-stats

72.0%
MAI-Thinking-1 self-reported llm-stats

72.0%
Qwen3-235B-A22B-Thinking-2507 self-reported llm-stats

71.9%
Qwen3 VL 235B A22B Thinking self-reported llm-stats

71.9%
Qwen3 VL 32B Thinking self-reported llm-stats

71.7%
Qwen3-235B-A22B-Instruct-2507 self-reported llm-stats

70.9%
Qwen3-Next-80B-A3B-Instruct self-reported llm-stats

70.3%
Qwen3 VL 32B Instruct self-reported llm-stats

70.2%
Qwen3 VL 30B A3B Thinking self-reported llm-stats

68.6%
Qwen3 VL 235B A22B Instruct self-reported llm-stats

67.7%
Qwen3 VL 4B Thinking self-reported llm-stats

67.3%
Qwen3 VL 30B A3B Instruct self-reported llm-stats

66.3%
Qwen3 VL 8B Instruct self-reported llm-stats

66.3%
Qwen3 VL 4B Instruct self-reported llm-stats

63.3%
Qwen3 VL 8B Thinking self-reported llm-stats

63.0%