벤치마크 상세
BFCL V4
함수 호출과 도구 실행 정확도: API 호출형 agent, 플러그인, 업무 자동화
공식 리더보드 보기
공식 Overall Accuracy를 그대로 사용합니다. FC와 Prompt 방식은 모델명에 남겼습니다. 공식 페이지 마지막 갱신은 2026-04-12이며 확인일과 평가 갱신일을 구분합니다.
모든 공개 결과
모델 검색 · 점수순확인 2026. 09. 07.원문 기준 2026. 04. 12.BFCL V4 · f7cf735 · bfcl-eval 2025.12.17% overall accuracy
109개 결과원문 ↗
| 순서 | 모델 / 실행 구성 | 점수 | 평가 조건·근거 |
|---|---|---|---|
| 1 | Claude-Opus-4-5-20251101 (FC)Anthropic | 77.47 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 2 | Claude-Sonnet-4-5-20250929 (FC)Anthropic | 73.24 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 3 | Gemini-3-Pro-Preview (Prompt)Google | 72.51 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 4 | GLM-4.6 (FC thinking)Zhipu AI | 72.38 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 5 | Grok-4-1-fast-reasoning (FC)xAI | 69.57 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 6 | Claude-Haiku-4-5-20251001 (FC)Anthropic | 68.7 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 7 | Gemini-3-Pro-Preview (FC)Google | 68.14 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 8 | o3-2025-04-16 (Prompt)OpenAI | 63.05 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 9 | Grok-4-0709 (Prompt)xAI | 62.97 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 10 | Grok-4-0709 (FC)xAI | 61.38 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 11 | Moonshotai-Kimi-K2-Instruct (FC)MoonshotAI | 59.06 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 12 | Grok-4-1-fast-non-reasoning (FC)xAI | 58.29 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 13 | Command A Reasoning (FC)Cohere | 57.06 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 14 | DeepSeek-V3.2-Exp (Prompt + Thinking)DeepSeek | 56.73 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 15 | Gemini-2.5-Flash (FC)Google | 56.24 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 16 | GPT-5.2-2025-12-11 (FC)OpenAI | 55.87 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 17 | GPT-5-mini-2025-08-07 (FC)OpenAI | 55.46 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 18 | xLAM-2-32b-fc-r (FC)Salesforce | 54.66 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 19 | DeepSeek-V3.2-Exp (FC)DeepSeek | 54.12 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 20 | GPT-4.1-2025-04-14 (FC)OpenAI | 53.96 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 21 | o4-mini-2025-04-16 (FC)OpenAI | 53.24 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 22 | xLAM-2-70b-fc-r (FC)Salesforce | 53.07 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 23 | Qwen3-235B-A22B-Instruct-2507 (Prompt)Qwen | 52.15 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 24 | GPT-5-nano-2025-08-07 (FC)OpenAI | 51.45 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 25 | Nanbeige4-3B-Thinking-2511 (FC)Nanbeige | 51.4 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 26 | Gemini-2.5-Flash (Prompt)Google | 50.9 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 27 | GPT-4.1-mini-2025-04-14 (FC)OpenAI | 50.45 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 28 | o4-mini-2025-04-16 (Prompt)OpenAI | 50.26 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 29 | Qwen3-32B (FC)Qwen | 48.71 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 30 | o3-2025-04-16 (FC)OpenAI | 48.56 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 31 | Qwen3-235B-A22B-Instruct-2507 (FC)Qwen | 47.99 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 32 | Nanbeige3.5-Pro-Thinking (FC)Nanbeige | 47.68 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 33 | Qwen3-32B (Prompt)Qwen | 46.78 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 34 | xLAM-2-8b-fc-r (FC)Salesforce | 46.68 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 35 | Command A (FC)Cohere | 46.49 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 36 | BitAgent-Bounty-8BBittensor | 46.23 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 37 | Arch-Agent-32Bkatanemo | 45.37 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 38 | GPT-5.2-2025-12-11 (Prompt)OpenAI | 45.27 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 39 | Qwen3-8B (FC)Qwen | 42.57 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 40 | ToolACE-2-8B (FC)Huawei Noah & USTC | 42.44 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 41 | Qwen3-30B-A3B-Instruct-2507 (FC)Qwen | 41.39 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 42 | xLAM-2-3b-fc-r (FC)Salesforce | 41.22 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 43 | Qwen3-14B (FC)Qwen | 41.03 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 44 | Qwen3-8B (Prompt)Qwen | 40.43 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 45 | GPT-4.1-2025-04-14 (Prompt)OpenAI | 39.38 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 46 | mistral-large-2411 (FC)Mistral AI | 38.37 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 47 | Qwen3-14B (Prompt)Qwen | 37.77 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 48 | Mistral-Medium-2505Mistral AI | 37.69 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 49 | Mistral-Medium-2505 (FC)Mistral AI | 37.56 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 50 | Llama-4-Maverick-17B-128E-Instruct-FP8 (FC)Meta | 37.29 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 51 | Mistral-small-2506 (FC)Mistral AI | 37.15 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 52 | Gemini-2.5-Flash-Lite (FC)Google | 36.87 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 53 | Qwen3-30B-A3B-Instruct-2507 (Prompt)Qwen | 36.7 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 54 | Qwen3-4B-Instruct-2507 (FC)Qwen | 35.68 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 55 | Qwen3-4B-Instruct-2507 (Prompt)Qwen | 35.52 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 56 | Arch-Agent-3Bkatanemo | 35.36 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 57 | Claude-Opus-4-5-20251101 (Prompt)Anthropic | 33.47 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 58 | GPT-4.1-nano-2025-04-14 (FC)OpenAI | 33.05 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 59 | Mistral-Small-2506 (Prompt)Mistral AI | 32.38 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 60 | Arch-Agent-1.5Bkatanemo | 32.14 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 61 | Command R7B (FC)Cohere | 32.07 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 62 | Llama-3.3-70B-Instruct (FC)Meta | 31.9 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 63 | mistral-large-2411 (Prompt)Mistral AI | 31.84 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 64 | Hammer2.1-7b (FC)MadeAgents | 31.67 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 65 | xLAM-2-1b-fc-r (FC)Salesforce | 30.44 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 66 | Gemma-3-12b-it (Prompt)Google | 30.43 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 67 | GPT-4.1-mini-2025-04-14 (Prompt)OpenAI | 29.73 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 68 | Hammer2.1-3b (FC)MadeAgents | 29.71 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 69 | Gemma-3-27b-it (Prompt)Google | 29.47 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 70 | Phi-4 (Prompt)Microsoft | 28.79 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 71 | Qwen3-1.7B (FC)Qwen | 28.41 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 72 | Llama-4-Scout-17B-16E-Instruct (FC)Meta | 28.13 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 73 | Gemini-2.5-Flash-Lite (Prompt)Google | 28.03 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 74 | CoALM-70BUIUC + Oumi | 27.99 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 75 | Hammer2.1-1.5b (FC)MadeAgents | 27.88 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 76 | palmyra-x-004 (FC)Writer | 27.87 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 77 | GPT-5-mini-2025-08-07 (Prompt)OpenAI | 27.83 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 78 | Open-Mistral-Nemo-2407 (FC)Mistral AI | 27.63 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 79 | GPT-5-nano-2025-08-07 (Prompt)OpenAI | 27.55 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 80 | Amazon-Nova-2-Lite-v1:0 (FC)Amazon | 27.1 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 81 | Granite-3.1-8B-Instruct (FC)IBM | 27.1 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 82 | Falcon3-10B-Instruct (FC)TII UAE | 27.01 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 83 | Granite-3.2-8B-Instruct (FC)IBM | 26.87 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 84 | CoALM-8BUIUC + Oumi | 26.81 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 85 | Llama-3.1-8B-Instruct (Prompt)Meta | 25.83 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 86 | MiniCPM3-4B-FC (FC)openbmb | 25.55 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 87 | Claude-Haiku-4-5-20251001 (Prompt)Anthropic | 25.26 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 88 | Amazon-Nova-Pro-v1:0 (FC)Amazon | 24.97 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 89 | Claude-Sonnet-4-5-20250929 (Prompt)Anthropic | 24.9 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 90 | GPT-4.1-nano-2025-04-14 (Prompt)OpenAI | 24.88 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 91 | Falcon3-7B-Instruct (FC)TII UAE | 24.03 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 92 | Qwen3-0.6B (FC)Qwen | 23.93 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 93 | Granite-20b-FunctionCalling (FC)IBM | 23.23 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 94 | Qwen3-0.6B (Prompt)Qwen | 22.38 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 95 | Amazon-Nova-Micro-v1:0 (FC)Amazon | 22.29 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 96 | RZN-T (Prompt)Phronetic AI | 22.25 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 97 | MiniCPM3-4B (Prompt)openbmb | 22.08 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 98 | Llama-3.2-3B-Instruct (FC)Meta | 21.95 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 99 | Bielik-11B-v2.3-Instruct (Prompt)SpeakLeash & ACK Cyfronet AGH | 21.9 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 100 | Hammer2.1-0.5b (FC)MadeAgents | 21.22 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 101 | Gemma-3-4b-it (Prompt)Google | 19.62 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 102 | Open-Mistral-Nemo-2407 (Prompt)Mistral AI | 19.31 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 103 | Granite-4.0-350m (FC)IBM | 18.98 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 104 | Falcon3-3B-Instruct (FC)TII UAE | 16.25 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 105 | Ministral-8B-Instruct-2410 (FC)Mistral AI | 11.1 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 106 | Falcon3-1B-Instruct (FC)TII UAE | 11.08 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 107 | Llama-3.2-1B-Instruct (FC)Meta | 10.82 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 108 | Llama-3.1-Nemotron-Ultra-253B-v1 (FC)NVIDIA | 10 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
| 109 | Gemma-3-1b-it (Prompt)Google | 7.17 | BFCL V4 · f7cf735 · bfcl-eval 2025.12.17 |
일치하는 모델이 없습니다. 검색어를 줄여 보세요.
공개된 숫자만 높은 순으로 표시합니다. 순서가 신뢰구간을 고려한 확정 순위는 아닙니다. 공식 Overall Accuracy를 그대로 사용합니다. FC와 Prompt 방식은 모델명에 남겼습니다. 공식 페이지 마지막 갱신은 2026-04-12이며 확인일과 평가 갱신일을 구분합니다.
읽는 법
전체 벤치마크Berkeley Function Calling Leaderboard. 함수 호출과 tool use 정확도를 봅니다.
챗봇이 아니라 실제 서비스를 조작하는 AI를 만들 때 중요합니다.