OpenGPT근거 기반 모델 선택
검색⌘K모델 선택
에이전트 벤치마크

Berkeley Function-Calling Leaderboard 에이전트 벤치마크

Berkeley Function-Calling Leaderboard의 표시된 버전과 환경에 연결된 공개 행 109개를 유지합니다.

과제
—
공개 행
109
출처 데이터 날짜
ID 검토일
과거 기록

근거 범위

BFCL V4 · commit f7cf735 · bfcl-eval 2025.12.17. official V4 cases; count not frozen in this snapshot. Function calling, web search, memory, multi-turn, hallucination, and format tests. Category-specific.

OpenGPT의 근거 처리 방식

게시자의 지표, 과제 범위, 환경, 버전, 구성과 날짜를 보존하며 벤치마크 간 종합 점수를 만들지 않습니다.

비교 범위

  • 높은 결과는 표시된 벤치마크 버전과 조건에만 적용됩니다.
  • 에이전트 결과는 기반 모델 하나가 아닌 전체 시스템을 반영합니다.
  • 비용, 신뢰성, 안전성 또는 지연 시간 누락값은 추정하지 않습니다.
  • OpenGPT includes all 109 rows from the publisher CSV snapshot. Every row shares commit f7cf735, bfcl-eval 2025.12.17, and the official V4 overall-accuracy composition; cost is the publisher estimate for the full benchmark.

BFCL V4 · commit f7cf735 · bfcl-eval 2025.12.17

공개 109개 중 24개를 표시합니다. 모든 결과는 출처 벤치마크와 구성에 연결됩니다.

  1. BFCL V4 · Claude-Opus-4-5-20251101 (FC)Anthropic · Claude-Opus-4-5-20251101 (FC)#1Overall accuracy: 77.47%BFCL V4 tool-use evaluation
  2. BFCL V4 · Claude-Sonnet-4-5-20250929 (FC)Anthropic · Claude-Sonnet-4-5-20250929 (FC)#2Overall accuracy: 73.24%BFCL V4 tool-use evaluation
  3. BFCL V4 · Gemini-3-Pro-Preview (Prompt)Google · Gemini-3-Pro-Preview (Prompt)#3Overall accuracy: 72.51%BFCL V4 tool-use evaluation
  4. BFCL V4 · GLM-4.6 (FC thinking)Zhipu AI · GLM-4.6 (FC thinking)#4Overall accuracy: 72.38%BFCL V4 tool-use evaluation
  5. BFCL V4 · Grok-4-1-fast-reasoning (FC)xAI · Grok-4-1-fast-reasoning (FC)#5Overall accuracy: 69.57%BFCL V4 tool-use evaluation
  6. BFCL V4 · Claude-Haiku-4-5-20251001 (FC)Anthropic · Claude-Haiku-4-5-20251001 (FC)#6Overall accuracy: 68.70%BFCL V4 tool-use evaluation
  7. BFCL V4 · Gemini-3-Pro-Preview (FC)Google · Gemini-3-Pro-Preview (FC)#7Overall accuracy: 68.14%BFCL V4 tool-use evaluation
  8. BFCL V4 · o3-2025-04-16 (Prompt)OpenAI · o3-2025-04-16 (Prompt)#8Overall accuracy: 63.05%BFCL V4 tool-use evaluation
  9. BFCL V4 · Grok-4-0709 (Prompt)xAI · Grok-4-0709 (Prompt)#9Overall accuracy: 62.97%BFCL V4 tool-use evaluation
  10. BFCL V4 · Grok-4-0709 (FC)xAI · Grok-4-0709 (FC)#10Overall accuracy: 61.38%BFCL V4 tool-use evaluation
  11. BFCL V4 · Moonshotai-Kimi-K2-Instruct (FC)MoonshotAI · Moonshotai-Kimi-K2-Instruct (FC)#11Overall accuracy: 59.06%BFCL V4 tool-use evaluation
  12. BFCL V4 · Grok-4-1-fast-non-reasoning (FC)xAI · Grok-4-1-fast-non-reasoning (FC)#12Overall accuracy: 58.29%BFCL V4 tool-use evaluation
  13. BFCL V4 · Command A Reasoning (FC)Cohere · Command A Reasoning (FC)#13Overall accuracy: 57.06%BFCL V4 tool-use evaluation
  14. BFCL V4 · DeepSeek-V3.2-Exp (Prompt + Thinking)DeepSeek · DeepSeek-V3.2-Exp (Prompt + Thinking)#14Overall accuracy: 56.73%BFCL V4 tool-use evaluation
  15. BFCL V4 · Gemini-2.5-Flash (FC)Google · Gemini-2.5-Flash (FC)#15Overall accuracy: 56.24%BFCL V4 tool-use evaluation
  16. BFCL V4 · GPT-5.2-2025-12-11 (FC)OpenAI · GPT-5.2-2025-12-11 (FC)#16Overall accuracy: 55.87%BFCL V4 tool-use evaluation
  17. BFCL V4 · GPT-5-mini-2025-08-07 (FC)OpenAI · GPT-5-mini-2025-08-07 (FC)#17Overall accuracy: 55.46%BFCL V4 tool-use evaluation
  18. BFCL V4 · xLAM-2-32b-fc-r (FC)Salesforce · xLAM-2-32b-fc-r (FC)#18Overall accuracy: 54.66%BFCL V4 tool-use evaluation
  19. BFCL V4 · DeepSeek-V3.2-Exp (FC)DeepSeek · DeepSeek-V3.2-Exp (FC)#19Overall accuracy: 54.12%BFCL V4 tool-use evaluation
  20. BFCL V4 · GPT-4.1-2025-04-14 (FC)OpenAI · GPT-4.1-2025-04-14 (FC)#20Overall accuracy: 53.96%BFCL V4 tool-use evaluation
  21. BFCL V4 · o4-mini-2025-04-16 (FC)OpenAI · o4-mini-2025-04-16 (FC)#21Overall accuracy: 53.24%BFCL V4 tool-use evaluation
  22. BFCL V4 · xLAM-2-70b-fc-r (FC)Salesforce · xLAM-2-70b-fc-r (FC)#22Overall accuracy: 53.07%BFCL V4 tool-use evaluation
  23. BFCL V4 · Qwen3-235B-A22B-Instruct-2507 (Prompt)Qwen · Qwen3-235B-A22B-Instruct-2507 (Prompt)#23Overall accuracy: 52.15%BFCL V4 tool-use evaluation
  24. BFCL V4 · GPT-5-nano-2025-08-07 (FC)OpenAI · GPT-5-nano-2025-08-07 (FC)#24Overall accuracy: 51.45%BFCL V4 tool-use evaluation