reasoning benchmark
BFCL-v3 Leaderboard
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.
Updated Aug 17, 2026
Models19
Model coverage19
MetricScore
EvidenceB
BFCL-v3 Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelZA | Score77.8% | Percentile100.0% | Participants19 | EvidenceC | Evaluated |
| Rank02 | ModelZA | Score76.4% | Percentile94.4% | Participants19 | EvidenceC | Evaluated |
| Rank03 | ModelME | Score74.4% | Percentile88.9% | Participants19 | EvidenceC | Evaluated |
| Rank04 | ModelMI | Score72.0% | Percentile83.3% | Participants19 | EvidenceC | Evaluated |
| Rank05 | ModelAC | Score72.0% | Percentile77.8% | Participants19 | EvidenceC | Evaluated |
| Rank06 | ModelAC | Score71.9% | Percentile72.2% | Participants19 | EvidenceC | Evaluated |
| Rank07 | ModelAC | Score71.9% | Percentile66.7% | Participants19 | EvidenceC | Evaluated |
| Rank08 | ModelAC | Score71.7% | Percentile61.1% | Participants19 | EvidenceC | Evaluated |
| Rank09 | ModelAC | Score70.9% | Percentile55.6% | Participants19 | EvidenceC | Evaluated |
| Rank10 | ModelAC | Score70.3% | Percentile50.0% | Participants19 | EvidenceC | Evaluated |
| Rank11 | ModelAC | Score70.2% | Percentile44.4% | Participants19 | EvidenceC | Evaluated |
| Rank12 | ModelAC | Score68.7% | Percentile38.9% | Participants19 | EvidenceC | Evaluated |
| Rank13 | ModelAC | Score68.6% | Percentile33.3% | Participants19 | EvidenceC | Evaluated |
| Rank14 | ModelAC | Score67.7% | Percentile27.8% | Participants19 | EvidenceC | Evaluated |
| Rank15 | ModelAC | Score67.3% | Percentile22.2% | Participants19 | EvidenceC | Evaluated |
| Rank16 | ModelAC | Score66.3% | Percentile16.7% | Participants19 | EvidenceC | Evaluated |
| Rank17 | ModelAC | Score66.3% | Percentile11.1% | Participants19 | EvidenceC | Evaluated |
| Rank18 | ModelAC | Score63.3% | Percentile5.6% | Participants19 | EvidenceC | Evaluated |
| Rank19 | ModelAC | Score63.0% | Percentile0.0% | Participants19 | EvidenceC | Evaluated |
BFCL-v3 Highlights
The leading models and scores on this benchmark.
BFCL-v3 Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for BFCL-v3
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
Ranking basisThis bfcl-v3 AI model leaderboard uses descending score in the benchmark's original unit. The leaderboard ranking keeps matched price and speed data separate from benchmark evidence.
- 01ZAGLM-4.5Zhipu AI
- Score
- 77.8%
- Price
- $0.60 input / $2.2 output per 1M tokens
Strengths
- Ranks #1 of 19 compared models
- 100th percentile on this benchmark
- C evidence result
Considerations
- This result measures BFCL-v3, not total model capability
- 02ZAGLM-4.5-AirZhipu AI
- Score
- 76.4%
- Price
- $0.20 input / $1.1 output per 1M tokens
Strengths
- Ranks #2 of 19 compared models
- 94th percentile on this benchmark
- C evidence result
Considerations
- This result measures BFCL-v3, not total model capability
- 03MELongCat-Flash-ThinkingMeituan
- Score
- 74.4%
- Speed
- Up to 100 tok/s via Meituan
Strengths
- Ranks #3 of 19 compared models
- 89th percentile on this benchmark
- C evidence result
Considerations
- This result measures BFCL-v3, not total model capability
- 04MIMAI-Thinking-1Microsoft
- Score
- 72.0%
Strengths
- Ranks #4 of 19 compared models
- 83th percentile on this benchmark
- C evidence result
Considerations
- This result measures BFCL-v3, not total model capability
- 05ACQwen3-Next-80B-A3B-ThinkingAlibaba Cloud / Qwen Team
- Score
- 72.0%
- Price
- $0.50 input / $6.0 output per 1M tokens
Strengths
- Ranks #5 of 19 compared models
- 78th percentile on this benchmark
- C evidence result
Considerations
- This result measures BFCL-v3, not total model capability
Selection summary
Best AI Models for BFCL-v3
GLM-4.5 currently leads BFCL-v3 with 77.8%. It is the top model on this specific benchmark, while the best LLM for the broader task should also be checked against other benchmarks, price and runtime.
Use this leaderboard with the supporting benchmark results and coverage details above. A leaderboard position summarizes the selected ranking signal; it does not replace workload-specific testing.
What is BFCL-v3?
What BFCL-v3 measures and how its scores work.
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.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- BFCL-v3
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- bfcl-v3|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about BFCL-v3.
Which model scores highest on BFCL-v3?
GLM-4.5 is currently ranked first with 77.8%.
What does BFCL-v3 measure?
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.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
19 model results are currently shown.
Does this benchmark affect the overall score?
Yes. This benchmark can contribute to the current LLMBoard capability score.