Benchmark Almanac

Berkeley Function Calling Leaderboard (BFCL)

Llama 3.1-405B 88.5% leads since July 2024
Llama-xLAM-2-70b-fc-r 78.5% leads since March 2025
Llama 3.1-405B 88.5% leads since July 2024
In use

Model emits function calls for single, parallel, multi-turn and agentic scenarios. Scored by AST and execution checks.

Live benchmark. Nova 2 pins results to a 2025 commit, so versions matter.

Released
February 2024
Kind
Function calling
Level
Everyday
Scored
Checked automatically
Answers
Agent task
Tasks
All public
Strongest model tested
Qwen3-235B-A22B-Thinking Jul 2025 #107 on the capability index
Last in a launch post
October 2026
Last model added
October 2026
Scores refreshed
November 12th, 2026
41
Models scored in lab cards
36
Launch posts list it, from 18 labs
31
System cards report it, from 11 labs

Llama 3.1-405B leads with 88.5%

Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.

Llama-xLAM-2-70b-fc-r leads with 78.5% across all independent runs

Best score per model across the official leaderboard, Epoch AI’s own runs and any board Epoch AI copied. Each bar and row names its source.

Llama 3.1-405B leads with 88.5%, claims included

Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model.

41 models. Scroll sideways for the rest.

  1. 88.5 Llama 3.1-405B
  2. 84.8 Llama 3.1-70B
  3. 78.3 Hunyuan-A13B-Instruct
  4. 77.8 GLM-4.5
  5. 76.4 GLM-4.5-Air
  6. 76.1 Llama 3.1-8B
  7. 74.4 LongCat-Flash-Thinking
  8. 73.7 Granite 4.1 30B
  9. 72.0 Qwen3-Next-80B-A3B
  10. 72.0 MAI-Thinking-1
  11. 71.9 Qwen3-235B-A22B-Thinking
  12. 71.9 Qwen3-VL-235B-A22B
  13. 71.8 Llama-3.3-Nemotron-Super-49B-v1.5
  14. 70.8 Qwen3-235B-A22B
  15. 70.5 GLM-Z1-32B-0414
  16. 70.4 Qwen3-14B
  17. 70.3 Phi-4 Mini
  18. 70.3 Qwen3-32B
  19. 70.3 Command A Reasoning
  20. 69.1 Qwen3-30B-A3B
  21. 68.8 Ring-1T
  22. 68.7 Qwen3-Coder-480B-A35B
  23. 68.1 Qwen3-8B
  24. 67.3 Nemotron 3.5 Bolt E4B
  25. 66.9 nvidia-nemotron-nano-9b-v2
  26. 65.9 Qwen3-4B
  27. 65.2 EXAONE 4.0 32B
  28. 65.0 ERNIE-4.5-21B-A3B-Thinking
  29. 64.7 Granite 4.0 H Small
  30. 64.4 Qwen3-Omni-30B-A3B
  31. 64.4 LFM2.5-8B-A1B
  32. 63.9 Solar Pro 2
  33. 63.8 Cohere Command A
  34. 63.7 Amazon Nova Premier
  35. 61.1 Nemotron 3 Nano 4B
  36. 59.7 Trinity Mini
  37. 57.9 ChatGLM3-6B
  38. 57.0 LFM2.5-1.2B-Thinking
  39. 56.6 Qwen3-1.7B
  40. 54.8 granite-4.0-1b
  41. 49.1 LFM2.5-1.2B-Instruct
  1. 1 Llama 3.1-405B 88.5%
  2. 2 Llama 3.1-70B 84.8%
  3. 3 Hunyuan-A13B-Instruct 78.3%
  4. 4 GLM-4.5 77.8%
  5. 5 GLM-4.5-Air 76.4%
  6. 6 Llama 3.1-8B 76.1%
  7. 7 LongCat-Flash-Thinking 74.4%
  8. 8 Granite 4.1 30B 73.7%
  9. 9 Qwen3-Next-80B-A3B 72%
  10. 10 MAI-Thinking-1 72%
  11. 11 Qwen3-235B-A22B-Thinking Jul 2025 71.9%
  12. 12 Qwen3-VL-235B-A22B 71.9%
  13. 13 Llama-3.3-Nemotron-Super-49B-v1.5 71.8%
  14. 14 Qwen3-235B-A22B 70.8%
  15. 15 GLM-Z1-32B-0414 70.5%
  16. 16 Qwen3-14B 70.4%
  17. 17 Phi-4 Mini 70.3%
  18. 18 Qwen3-32B 70.3%
  19. 19 Command A Reasoning 70.3%
  20. 20 Qwen3-30B-A3B 69.1%
  21. 21 Ring-1T 68.8%
  22. 22 Qwen3-Coder-480B-A35B 68.7%
  23. 23 Qwen3-8B 68.1%
  24. 24 Nemotron 3.5 Bolt E4B 67.3%
  25. 25 nvidia-nemotron-nano-9b-v2 66.9%
  26. 26 Qwen3-4B 65.9%
  27. 27 EXAONE 4.0 32B 65.2%
  28. 28 ERNIE-4.5-21B-A3B-Thinking 65%
  29. 29 Granite 4.0 H Small 64.7%
  30. 30 Qwen3-Omni-30B-A3B 64.4%
  31. 31 LFM2.5-8B-A1B 64.4%
  32. 32 Solar Pro 2 63.9%
  33. 33 Cohere Command A 63.8%
  34. 34 Amazon Nova Premier 63.7%
  35. 35 Nemotron 3 Nano 4B 61.1%
  36. 36 Trinity Mini 59.7%
  37. 37 ChatGLM3-6B 57.9%
  38. 38 LFM2.5-1.2B-Thinking 57%
  39. 39 Qwen3-1.7B 56.6%
  40. 40 granite-4.0-1b 54.8%
  41. 41 LFM2.5-1.2B-Instruct 49.1%

Source: Labs’ model cards

67 models. Scroll sideways for the rest.

  1. 78.5 Llama-xLAM-2-70b-fc-r
  2. 76.4 xLAM-2-32b-fc-r
  3. 72.0 Llama-xLAM-2-8b-fc-r
  4. 71.7 GPT-4o
  5. 70.3 GPT-4.5
  6. 69.3 Qwen3-32B
  7. 68.9 GPT-4.1
  8. 68.0 Qwen3-14B
  9. 67.9 Qwen3-235B-A22B
  10. 67.8 o1
  11. 66.3 Qwen3-8B
  12. 65.7 Qwen3-30B-A3B
  13. 65.2 GPT-4.1 mini
  14. 64.8 xLAM-2-3b-fc-r
  15. 64.7 DeepSeek-V3
  16. 64.6 o3-mini
  17. 64.4 Cohere Command A
  18. 64.0 GPT-4o mini
  19. 63.8 DeepSeek-R1
  20. 63.4 Qwen3-4B
  21. 63.3 Grok 3
  22. 61.2 Gemini 2.0 Flash
  23. 60.8 Amazon Nova Pro
  24. 59.9 Gemini 2.5 Pro
  25. 59.1 Gemma 3 27B
  26. 58.7 Gemini 2.0 Flash Thinking
  27. 58.6 Claude 3.7 Sonnet
  28. 58.3 QwQ-32B
  29. 58.2 xLAM-2-1b-fc-r
  30. 58.0 Mistral Large 2
  31. 57.7 Mistral Small 3.1
  32. 57.0 Claude 3 Opus
  33. 56.9 Claude 3.5 Sonnet (October 2024)
  34. 56.8 Palmyra X4
  35. 56.7 Llama-3.1-Nemotron-Ultra-253B-v1
  36. 56.4 Gemini 2.0 Flash-Lite
  37. 55.9 Amazon Nova Lite
  38. 54.7 Qwen3-1.7B
  39. 53.7 Claude 3.5 Haiku
  40. 53.6 Llama 3.1-70B
  41. 53.0 Grok-3 mini
  42. 53.0 GPT-4.1 nano
  43. 52.9 Llama 4 Maverick
  44. 52.2 Llama 3.3 70B Instruct
  45. 51.8 Amazon Nova Micro
  46. 51.7 Command R7B
  47. 51.5 Ministral 8B
  48. 51.5 Mistral NeMo
  49. 50.6 Gemma 3 12B
  50. 50.3 MiniCPM3-4B
  51. 49.6 Llama 3.1-8B
  52. 48.9 Granite-20B-FunctionCalling
  53. 48.0 Command R+
  54. 47.3 Qwen3-0.6B
  55. 47.0 Phi-4
  56. 45.9 Llama 3.2 3B
  57. 45.4 Llama 4 Scout
  58. 45.1 Falcon3-10B-Instruct
  59. 43.6 Falcon3-7B-Instruct
  60. 39.6 Gemma 3 4B
  61. 39.2 DBRX
  62. 39.1 GLM-4-9B-Chat
  63. 32.5 Falcon3-3B-Instruct
  64. 31.3 Phi-4 Mini
  65. 21.9 Llama 3.2 1B
  66. 19.7 Falcon3-1B-Instruct
  67. 16.3 Gemma 3 1B
  1. 1 Llama-xLAM-2-70b-fc-r 78.5%
  2. 2 xLAM-2-32b-fc-r 76.4%
  3. 3 Llama-xLAM-2-8b-fc-r 72%
  4. 4 GPT-4o Nov 2024 71.7%
  5. 5 GPT-4.5 70.3%
  6. 6 Qwen3-32B 69.3%
  7. 7 GPT-4.1 68.9%
  8. 8 Qwen3-14B 68%
  9. 9 Qwen3-235B-A22B 67.9%
  10. 10 o1 67.8%
  11. 11 Qwen3-8B 66.3%
  12. 12 Qwen3-30B-A3B 65.7%
  13. 13 GPT-4.1 mini 65.2%
  14. 14 xLAM-2-3b-fc-r 64.8%
  15. 15 DeepSeek-V3 Mar 2025 64.7%
  16. 16 o3-mini 64.6%
  17. 17 Cohere Command A 64.4%
  18. 18 GPT-4o mini 64%
  19. 19 DeepSeek-R1 May 2025 63.8%
  20. 20 Qwen3-4B 63.4%
  21. 21 Grok 3 63.3%
  22. 22 Gemini 2.0 Flash Feb 2025 61.2%
  23. 23 Amazon Nova Pro 60.8%
  24. 24 Gemini 2.5 Pro May 2025 59.9%
  25. 25 Gemma 3 27B 59.1%
  26. 26 Gemini 2.0 Flash Thinking Jan 2025 58.7%
  27. 27 Claude 3.7 Sonnet 58.6%
  28. 28 QwQ-32B 58.3%
  29. 29 xLAM-2-1b-fc-r 58.2%
  30. 30 Mistral Large 2 Nov 2024 58%
  31. 31 Mistral Small 3.1 57.7%
  32. 32 Claude 3 Opus 57%
  33. 33 Claude 3.5 Sonnet (October 2024) 56.9%
  34. 34 Palmyra X4 56.8%
  35. 35 Llama-3.1-Nemotron-Ultra-253B-v1 56.7%
  36. 36 Gemini 2.0 Flash-Lite 56.4%
  37. 37 Amazon Nova Lite 55.9%
  38. 38 Qwen3-1.7B 54.7%
  39. 39 Claude 3.5 Haiku 53.7%
  40. 40 Llama 3.1-70B 53.6%
  41. 41 Grok-3 mini 53%
  42. 42 GPT-4.1 nano 53%
  43. 43 Llama 4 Maverick 52.9%
  44. 44 Llama 3.3 70B Instruct 52.2%
  45. 45 Amazon Nova Micro 51.8%
  46. 46 Command R7B 51.7%
  47. 47 Ministral 8B 51.5%
  48. 48 Mistral NeMo 51.5%
  49. 49 Gemma 3 12B 50.6%
  50. 50 MiniCPM3-4B 50.3%
  51. 51 Llama 3.1-8B 49.6%
  52. 52 Granite-20B-FunctionCalling 48.9%
  53. 53 Command R+ 48%
  54. 54 Qwen3-0.6B 47.3%
  55. 55 Phi-4 47%
  56. 56 Llama 3.2 3B 45.9%
  57. 57 Llama 4 Scout 45.4%
  58. 58 Falcon3-10B-Instruct 45.1%
  59. 59 Falcon3-7B-Instruct 43.6%
  60. 60 Gemma 3 4B 39.6%
  61. 61 DBRX 39.2%
  62. 62 GLM-4-9B-Chat 39.1%
  63. 63 Falcon3-3B-Instruct 32.5%
  64. 64 Phi-4 Mini 31.3%
  65. 65 Llama 3.2 1B 21.9%
  66. 66 Falcon3-1B-Instruct 19.7%
  67. 67 Gemma 3 1B 16.3%

Source: Official Berkeley Function Calling Leaderboard (BFCL) leaderboard

97 models. Scroll sideways for the rest.

  1. 88.5 Llama 3.1-405B
  2. 84.8 Llama 3.1-70B
  3. 78.5 Llama-xLAM-2-70b-fc-r
  4. 78.3 Hunyuan-A13B-Instruct
  5. 77.8 GLM-4.5
  6. 76.4 xLAM-2-32b-fc-r
  7. 76.4 GLM-4.5-Air
  8. 76.1 Llama 3.1-8B
  9. 74.4 LongCat-Flash-Thinking
  10. 73.7 Granite 4.1 30B
  11. 72.0 Llama-xLAM-2-8b-fc-r
  12. 72.0 Qwen3-Next-80B-A3B
  13. 72.0 MAI-Thinking-1
  14. 71.9 Qwen3-235B-A22B-Thinking
  15. 71.9 Qwen3-VL-235B-A22B
  16. 71.8 Llama-3.3-Nemotron-Super-49B-v1.5
  17. 71.7 GPT-4o
  18. 70.8 Qwen3-235B-A22B
  19. 70.5 GLM-Z1-32B-0414
  20. 70.4 Qwen3-14B
  21. 70.3 GPT-4.5
  22. 70.3 Phi-4 Mini
  23. 70.3 Qwen3-32B
  24. 70.3 Command A Reasoning
  25. 69.1 Qwen3-30B-A3B
  26. 68.9 GPT-4.1
  27. 68.8 Ring-1T
  28. 68.7 Qwen3-Coder-480B-A35B
  29. 68.1 Qwen3-8B
  30. 67.8 o1
  31. 67.3 Nemotron 3.5 Bolt E4B
  32. 66.9 nvidia-nemotron-nano-9b-v2
  33. 65.9 Qwen3-4B
  34. 65.2 EXAONE 4.0 32B
  35. 65.2 GPT-4.1 mini
  36. 65.0 ERNIE-4.5-21B-A3B-Thinking
  37. 64.8 xLAM-2-3b-fc-r
  38. 64.7 DeepSeek-V3
  39. 64.7 Granite 4.0 H Small
  40. 64.6 o3-mini
  41. 64.4 Qwen3-Omni-30B-A3B
  42. 64.4 Cohere Command A
  43. 64.4 LFM2.5-8B-A1B
  44. 64.0 GPT-4o mini
  45. 63.9 Solar Pro 2
  46. 63.8 DeepSeek-R1
  47. 63.7 Amazon Nova Premier
  48. 63.3 Grok 3
  49. 61.2 Gemini 2.0 Flash
  50. 61.1 Nemotron 3 Nano 4B
  51. 60.8 Amazon Nova Pro
  52. 59.9 Gemini 2.5 Pro
  53. 59.7 Trinity Mini
  54. 59.1 Gemma 3 27B
  55. 58.7 Gemini 2.0 Flash Thinking
  56. 58.6 Claude 3.7 Sonnet
  57. 58.3 QwQ-32B
  58. 58.2 xLAM-2-1b-fc-r
  59. 58.0 Mistral Large 2
  60. 57.9 ChatGLM3-6B
  61. 57.7 Mistral Small 3.1
  62. 57.0 LFM2.5-1.2B-Thinking
  63. 57.0 Claude 3 Opus
  64. 56.9 Claude 3.5 Sonnet (October 2024)
  65. 56.8 Palmyra X4
  66. 56.7 Llama-3.1-Nemotron-Ultra-253B-v1
  67. 56.6 Qwen3-1.7B
  68. 56.4 Gemini 2.0 Flash-Lite
  69. 55.9 Amazon Nova Lite
  70. 54.8 granite-4.0-1b
  71. 53.7 Claude 3.5 Haiku
  72. 53.0 Grok-3 mini
  73. 53.0 GPT-4.1 nano
  74. 52.9 Llama 4 Maverick
  75. 52.2 Llama 3.3 70B Instruct
  76. 51.8 Amazon Nova Micro
  77. 51.7 Command R7B
  78. 51.5 Ministral 8B
  79. 51.5 Mistral NeMo
  80. 50.6 Gemma 3 12B
  81. 50.3 MiniCPM3-4B
  82. 49.1 LFM2.5-1.2B-Instruct
  83. 48.9 Granite-20B-FunctionCalling
  84. 48.0 Command R+
  85. 47.3 Qwen3-0.6B
  86. 47.0 Phi-4
  87. 45.9 Llama 3.2 3B
  88. 45.4 Llama 4 Scout
  89. 45.1 Falcon3-10B-Instruct
  90. 43.6 Falcon3-7B-Instruct
  91. 39.6 Gemma 3 4B
  92. 39.2 DBRX
  93. 39.1 GLM-4-9B-Chat
  94. 32.5 Falcon3-3B-Instruct
  95. 21.9 Llama 3.2 1B
  96. 19.7 Falcon3-1B-Instruct
  97. 16.3 Gemma 3 1B
  1. 1 Llama 3.1-405B 88.5%
  2. 2 Llama 3.1-70B 84.8%
  3. 3 Llama-xLAM-2-70b-fc-r 78.5%
  4. 4 Hunyuan-A13B-Instruct 78.3%
  5. 5 GLM-4.5 77.8%
  6. 6 xLAM-2-32b-fc-r 76.4%
  7. 7 GLM-4.5-Air 76.4%
  8. 8 Llama 3.1-8B 76.1%
  9. 9 LongCat-Flash-Thinking 74.4%
  10. 10 Granite 4.1 30B 73.7%
  11. 11 Llama-xLAM-2-8b-fc-r 72%
  12. 12 Qwen3-Next-80B-A3B 72%
  13. 13 MAI-Thinking-1 72%
  14. 14 Qwen3-235B-A22B-Thinking Jul 2025 71.9%
  15. 15 Qwen3-VL-235B-A22B 71.9%
  16. 16 Llama-3.3-Nemotron-Super-49B-v1.5 71.8%
  17. 17 GPT-4o Nov 2024 71.7%
  18. 18 Qwen3-235B-A22B 70.8%
  19. 19 GLM-Z1-32B-0414 70.5%
  20. 20 Qwen3-14B 70.4%
  21. 21 GPT-4.5 70.3%
  22. 22 Phi-4 Mini 70.3%
  23. 23 Qwen3-32B 70.3%
  24. 24 Command A Reasoning 70.3%
  25. 25 Qwen3-30B-A3B 69.1%
  26. 26 GPT-4.1 68.9%
  27. 27 Ring-1T 68.8%
  28. 28 Qwen3-Coder-480B-A35B 68.7%
  29. 29 Qwen3-8B 68.1%
  30. 30 o1 67.8%
  31. 31 Nemotron 3.5 Bolt E4B 67.3%
  32. 32 nvidia-nemotron-nano-9b-v2 66.9%
  33. 33 Qwen3-4B 65.9%
  34. 34 EXAONE 4.0 32B 65.2%
  35. 35 GPT-4.1 mini 65.2%
  36. 36 ERNIE-4.5-21B-A3B-Thinking 65%
  37. 37 xLAM-2-3b-fc-r 64.8%
  38. 38 DeepSeek-V3 Mar 2025 64.7%
  39. 39 Granite 4.0 H Small 64.7%
  40. 40 o3-mini 64.6%
  41. 41 Qwen3-Omni-30B-A3B 64.4%
  42. 42 Cohere Command A 64.4%
  43. 43 LFM2.5-8B-A1B 64.4%
  44. 44 GPT-4o mini 64%
  45. 45 Solar Pro 2 63.9%
  46. 46 DeepSeek-R1 May 2025 63.8%
  47. 47 Amazon Nova Premier 63.7%
  48. 48 Grok 3 63.3%
  49. 49 Gemini 2.0 Flash Feb 2025 61.2%
  50. 50 Nemotron 3 Nano 4B 61.1%
  51. 51 Amazon Nova Pro 60.8%
  52. 52 Gemini 2.5 Pro May 2025 59.9%
  53. 53 Trinity Mini 59.7%
  54. 54 Gemma 3 27B 59.1%
  55. 55 Gemini 2.0 Flash Thinking Jan 2025 58.7%
  56. 56 Claude 3.7 Sonnet 58.6%
  57. 57 QwQ-32B 58.3%
  58. 58 xLAM-2-1b-fc-r 58.2%
  59. 59 Mistral Large 2 Nov 2024 58%
  60. 60 ChatGLM3-6B 57.9%
  61. 61 Mistral Small 3.1 57.7%
  62. 62 LFM2.5-1.2B-Thinking 57%
  63. 63 Claude 3 Opus 57%
  64. 64 Claude 3.5 Sonnet (October 2024) 56.9%
  65. 65 Palmyra X4 56.8%
  66. 66 Llama-3.1-Nemotron-Ultra-253B-v1 56.7%
  67. 67 Qwen3-1.7B 56.6%
  68. 68 Gemini 2.0 Flash-Lite 56.4%
  69. 69 Amazon Nova Lite 55.9%
  70. 70 granite-4.0-1b 54.8%
  71. 71 Claude 3.5 Haiku 53.7%
  72. 72 Grok-3 mini 53%
  73. 73 GPT-4.1 nano 53%
  74. 74 Llama 4 Maverick 52.9%
  75. 75 Llama 3.3 70B Instruct 52.2%
  76. 76 Amazon Nova Micro 51.8%
  77. 77 Command R7B 51.7%
  78. 78 Ministral 8B 51.5%
  79. 79 Mistral NeMo 51.5%
  80. 80 Gemma 3 12B 50.6%
  81. 81 MiniCPM3-4B 50.3%
  82. 82 LFM2.5-1.2B-Instruct 49.1%
  83. 83 Granite-20B-FunctionCalling 48.9%
  84. 84 Command R+ 48%
  85. 85 Qwen3-0.6B 47.3%
  86. 86 Phi-4 47%
  87. 87 Llama 3.2 3B 45.9%
  88. 88 Llama 4 Scout 45.4%
  89. 89 Falcon3-10B-Instruct 45.1%
  90. 90 Falcon3-7B-Instruct 43.6%
  91. 91 Gemma 3 4B 39.6%
  92. 92 DBRX 39.2%
  93. 93 GLM-4-9B-Chat 39.1%
  94. 94 Falcon3-3B-Instruct 32.5%
  95. 95 Llama 3.2 1B 21.9%
  96. 96 Falcon3-1B-Instruct 19.7%
  97. 97 Gemma 3 1B 16.3%

Sources: Labs’ model cards and official Berkeley Function Calling Leaderboard (BFCL) leaderboard

#ModelScoreSettingDateSource
1 Llama 3.1-405B 88.5% Jul 2024 Lab model card
2 Llama 3.1-70B 84.8% Jul 2024 Lab model card
3 Hunyuan-A13B-Instruct 78.3% Jun 2025 Lab model card
4 GLM-4.5 77.8% Jul 2025 Lab model card
5 GLM-4.5-Air 76.4% Aug 2025 Lab model card
6 Llama 3.1-8B 76.1% Jul 2024 Lab model card
7 LongCat-Flash-Thinking 74.4% Sep 2025 Lab model card
8 Granite 4.1 30B 73.7% Apr 2026 Lab model card
9 Qwen3-Next-80B-A3B 72% Sep 2025 Lab model card
10 MAI-Thinking-1 72% Jun 2026 Lab model card
11 Qwen3-235B-A22B-Thinking Jul 2025 71.9% Jul 2025 Lab model card
12 Qwen3-VL-235B-A22B 71.9% Sep 2025 Lab model card
13 Llama-3.3-Nemotron-Super-49B-v1.5 71.8% Jul 2025 Lab model card
14 Qwen3-235B-A22B 70.8% Apr 2025 Lab model card
15 GLM-Z1-32B-0414 70.5% Apr 2025 Lab model card
16 Qwen3-14B 70.4% May 2025 Lab model card
17 Phi-4 Mini 70.3% Mar 2025 Lab model card
18 Qwen3-32B 70.3% May 2025 Lab model card
19 Command A Reasoning 70.3% Aug 2025 Lab model card
20 Qwen3-30B-A3B 69.1% May 2025 Lab model card
21 Ring-1T 68.8% Oct 2025 Lab model card
22 Qwen3-Coder-480B-A35B 68.7% Jul 2025 Lab model card
23 Qwen3-8B 68.1% May 2025 Lab model card
24 Nemotron 3.5 Bolt E4B 67.3% Oct 2026 Lab model card
25 nvidia-nemotron-nano-9b-v2 66.9% Aug 2025 Lab model card
26 Qwen3-4B 65.9% May 2025 Lab model card
27 EXAONE 4.0 32B 65.2% Jul 2025 Lab model card
28 ERNIE-4.5-21B-A3B-Thinking 65% Sep 2025 Lab model card
29 Granite 4.0 H Small 64.7% Oct 2025 Lab model card
30 Qwen3-Omni-30B-A3B 64.4% Sep 2025 Lab model card
31 LFM2.5-8B-A1B 64.4% May 2026 Lab model card
32 Solar Pro 2 63.9% Jul 2025 Lab model card
33 Cohere Command A 63.8% Apr 2025 Lab model card
34 Amazon Nova Premier 63.7% Apr 2025 Lab model card
35 Nemotron 3 Nano 4B 61.1% Mar 2026 Lab model card
36 Trinity Mini 59.7% Dec 2025 Lab model card
37 ChatGLM3-6B 57.9% Jun 2024 Lab model card
38 LFM2.5-1.2B-Thinking 57% Jan 2026 Lab model card
39 Qwen3-1.7B 56.6% May 2025 Lab model card
40 granite-4.0-1b 54.8% Oct 2025 Lab model card
41 LFM2.5-1.2B-Instruct 49.1% Jan 2026 Lab model card
#ModelScoreSettingDateSource
1 Llama-xLAM-2-70b-fc-r 78.5% native function calling Mar 2025 Official leaderboard
2 xLAM-2-32b-fc-r 76.4% native function calling Mar 2025 Official leaderboard
3 Llama-xLAM-2-8b-fc-r 72% native function calling Mar 2025 Official leaderboard
4 GPT-4o Nov 2024 71.7% native function calling Nov 2024 Official leaderboard
5 GPT-4.5 70.3% native function calling Feb 2025 Official leaderboard
6 Qwen3-32B 69.3% native function calling Apr 2025 Official leaderboard
7 GPT-4.1 68.9% native function calling Apr 2025 Official leaderboard
8 Qwen3-14B 68% native function calling Apr 2025 Official leaderboard
9 Qwen3-235B-A22B 67.9% native function calling Apr 2025 Official leaderboard
10 o1 67.8% prompted Dec 2024 Official leaderboard
11 Qwen3-8B 66.3% native function calling Apr 2025 Official leaderboard
12 Qwen3-30B-A3B 65.7% native function calling Apr 2025 Official leaderboard
13 GPT-4.1 mini 65.2% native function calling Apr 2025 Official leaderboard
14 xLAM-2-3b-fc-r 64.8% native function calling Mar 2025 Official leaderboard
15 DeepSeek-V3 Mar 2025 64.7% native function calling Mar 2025 Official leaderboard
16 o3-mini 64.6% prompted Jan 2025 Official leaderboard
17 Cohere Command A 64.4% native function calling Mar 2025 Official leaderboard
18 GPT-4o mini 64% native function calling Jul 2024 Official leaderboard
19 DeepSeek-R1 May 2025 63.8% native function calling May 2025 Official leaderboard
20 Qwen3-4B 63.4% native function calling Apr 2025 Official leaderboard
21 Grok 3 63.3% native function calling Apr 2025 Official leaderboard
22 Gemini 2.0 Flash Feb 2025 61.2% native function calling Feb 2025 Official leaderboard
23 Amazon Nova Pro 60.8% native function calling Dec 2024 Official leaderboard
24 Gemini 2.5 Pro May 2025 59.9% native function calling May 2025 Official leaderboard
25 Gemma 3 27B 59.1% prompted Mar 2025 Official leaderboard
26 Gemini 2.0 Flash Thinking Jan 2025 58.7% prompted Jan 2025 Official leaderboard
27 Claude 3.7 Sonnet 58.6% native function calling Feb 2025 Official leaderboard
28 QwQ-32B 58.3% native function calling Mar 2025 Official leaderboard
29 xLAM-2-1b-fc-r 58.2% native function calling Mar 2025 Official leaderboard
30 Mistral Large 2 Nov 2024 58% native function calling Nov 2024 Official leaderboard
31 Mistral Small 3.1 57.7% native function calling Mar 2025 Official leaderboard
32 Claude 3 Opus 57% native function calling Feb 2024 Official leaderboard
33 Claude 3.5 Sonnet (October 2024) 56.9% native function calling Oct 2024 Official leaderboard
34 Palmyra X4 56.8% native function calling Oct 2024 Official leaderboard
35 Llama-3.1-Nemotron-Ultra-253B-v1 56.7% native function calling Apr 2025 Official leaderboard
36 Gemini 2.0 Flash-Lite 56.4% prompted Feb 2025 Official leaderboard
37 Amazon Nova Lite 55.9% native function calling Dec 2024 Official leaderboard
38 Qwen3-1.7B 54.7% native function calling Apr 2025 Official leaderboard
39 Claude 3.5 Haiku 53.7% prompted Oct 2024 Official leaderboard
40 Llama 3.1-70B 53.6% prompted Jul 2024 Official leaderboard
41 Grok-3 mini 53% native function calling Apr 2025 Official leaderboard
42 GPT-4.1 nano 53% native function calling Apr 2025 Official leaderboard
43 Llama 4 Maverick 52.9% native function calling Apr 2025 Official leaderboard
44 Llama 3.3 70B Instruct 52.2% native function calling Dec 2024 Official leaderboard
45 Amazon Nova Micro 51.8% native function calling Dec 2024 Official leaderboard
46 Command R7B 51.7% native function calling Dec 2024 Official leaderboard
47 Ministral 8B 51.5% native function calling Oct 2024 Official leaderboard
48 Mistral NeMo 51.5% native function calling Jul 2024 Official leaderboard
49 Gemma 3 12B 50.6% prompted Mar 2025 Official leaderboard
50 MiniCPM3-4B 50.3% native function calling Sep 2024 Official leaderboard
51 Llama 3.1-8B 49.6% prompted Jul 2024 Official leaderboard
52 Granite-20B-FunctionCalling 48.9% native function calling Jul 2024 Official leaderboard
53 Command R+ 48% native function calling Aug 2024 Official leaderboard
54 Qwen3-0.6B 47.3% native function calling Apr 2025 Official leaderboard
55 Phi-4 47% prompted Dec 2024 Official leaderboard
56 Llama 3.2 3B 45.9% native function calling Sep 2024 Official leaderboard
57 Llama 4 Scout 45.4% native function calling Apr 2025 Official leaderboard
58 Falcon3-10B-Instruct 45.1% native function calling Dec 2024 Official leaderboard
59 Falcon3-7B-Instruct 43.6% native function calling Dec 2024 Official leaderboard
60 Gemma 3 4B 39.6% prompted Mar 2025 Official leaderboard
61 DBRX 39.2% prompted Mar 2024 Official leaderboard
62 GLM-4-9B-Chat 39.1% native function calling Jun 2024 Official leaderboard
63 Falcon3-3B-Instruct 32.5% native function calling Dec 2024 Official leaderboard
64 Phi-4 Mini 31.3% prompted Feb 2025 Official leaderboard
65 Llama 3.2 1B 21.9% native function calling Sep 2024 Official leaderboard
66 Falcon3-1B-Instruct 19.7% native function calling Dec 2024 Official leaderboard
67 Gemma 3 1B 16.3% prompted Mar 2025 Official leaderboard
#ModelScoreSettingDateSource
1 Llama 3.1-405B 88.5% Jul 2024 Lab model card
2 Llama 3.1-70B 84.8% Jul 2024 Lab model card
3 Llama-xLAM-2-70b-fc-r 78.5% native function calling Mar 2025 Official leaderboard
4 Hunyuan-A13B-Instruct 78.3% Jun 2025 Lab model card
5 GLM-4.5 77.8% Jul 2025 Lab model card
6 xLAM-2-32b-fc-r 76.4% native function calling Mar 2025 Official leaderboard
7 GLM-4.5-Air 76.4% Aug 2025 Lab model card
8 Llama 3.1-8B 76.1% Jul 2024 Lab model card
9 LongCat-Flash-Thinking 74.4% Sep 2025 Lab model card
10 Granite 4.1 30B 73.7% Apr 2026 Lab model card
11 Llama-xLAM-2-8b-fc-r 72% native function calling Mar 2025 Official leaderboard
12 Qwen3-Next-80B-A3B 72% Sep 2025 Lab model card
13 MAI-Thinking-1 72% Jun 2026 Lab model card
14 Qwen3-235B-A22B-Thinking Jul 2025 71.9% Jul 2025 Lab model card
15 Qwen3-VL-235B-A22B 71.9% Sep 2025 Lab model card
16 Llama-3.3-Nemotron-Super-49B-v1.5 71.8% Jul 2025 Lab model card
17 GPT-4o Nov 2024 71.7% native function calling Nov 2024 Official leaderboard
18 Qwen3-235B-A22B 70.8% Apr 2025 Lab model card
19 GLM-Z1-32B-0414 70.5% Apr 2025 Lab model card
20 Qwen3-14B 70.4% May 2025 Lab model card
21 GPT-4.5 70.3% native function calling Feb 2025 Official leaderboard
22 Phi-4 Mini 70.3% Mar 2025 Lab model card
23 Qwen3-32B 70.3% May 2025 Lab model card
24 Command A Reasoning 70.3% Aug 2025 Lab model card
25 Qwen3-30B-A3B 69.1% May 2025 Lab model card
26 GPT-4.1 68.9% native function calling Apr 2025 Official leaderboard
27 Ring-1T 68.8% Oct 2025 Lab model card
28 Qwen3-Coder-480B-A35B 68.7% Jul 2025 Lab model card
29 Qwen3-8B 68.1% May 2025 Lab model card
30 o1 67.8% prompted Dec 2024 Official leaderboard
31 Nemotron 3.5 Bolt E4B 67.3% Oct 2026 Lab model card
32 nvidia-nemotron-nano-9b-v2 66.9% Aug 2025 Lab model card
33 Qwen3-4B 65.9% May 2025 Lab model card
34 EXAONE 4.0 32B 65.2% Jul 2025 Lab model card
35 GPT-4.1 mini 65.2% native function calling Apr 2025 Official leaderboard
36 ERNIE-4.5-21B-A3B-Thinking 65% Sep 2025 Lab model card
37 xLAM-2-3b-fc-r 64.8% native function calling Mar 2025 Official leaderboard
38 DeepSeek-V3 Mar 2025 64.7% native function calling Mar 2025 Official leaderboard
39 Granite 4.0 H Small 64.7% Oct 2025 Lab model card
40 o3-mini 64.6% prompted Jan 2025 Official leaderboard
41 Qwen3-Omni-30B-A3B 64.4% Sep 2025 Lab model card
42 Cohere Command A 64.4% native function calling Mar 2025 Official leaderboard
43 LFM2.5-8B-A1B 64.4% May 2026 Lab model card
44 GPT-4o mini 64% native function calling Jul 2024 Official leaderboard
45 Solar Pro 2 63.9% Jul 2025 Lab model card
46 DeepSeek-R1 May 2025 63.8% native function calling May 2025 Official leaderboard
47 Amazon Nova Premier 63.7% Apr 2025 Lab model card
48 Grok 3 63.3% native function calling Apr 2025 Official leaderboard
49 Gemini 2.0 Flash Feb 2025 61.2% native function calling Feb 2025 Official leaderboard
50 Nemotron 3 Nano 4B 61.1% Mar 2026 Lab model card
51 Amazon Nova Pro 60.8% native function calling Dec 2024 Official leaderboard
52 Gemini 2.5 Pro May 2025 59.9% native function calling May 2025 Official leaderboard
53 Trinity Mini 59.7% Dec 2025 Lab model card
54 Gemma 3 27B 59.1% prompted Mar 2025 Official leaderboard
55 Gemini 2.0 Flash Thinking Jan 2025 58.7% prompted Jan 2025 Official leaderboard
56 Claude 3.7 Sonnet 58.6% native function calling Feb 2025 Official leaderboard
57 QwQ-32B 58.3% native function calling Mar 2025 Official leaderboard
58 xLAM-2-1b-fc-r 58.2% native function calling Mar 2025 Official leaderboard
59 Mistral Large 2 Nov 2024 58% native function calling Nov 2024 Official leaderboard
60 ChatGLM3-6B 57.9% Jun 2024 Lab model card
61 Mistral Small 3.1 57.7% native function calling Mar 2025 Official leaderboard
62 LFM2.5-1.2B-Thinking 57% Jan 2026 Lab model card
63 Claude 3 Opus 57% native function calling Feb 2024 Official leaderboard
64 Claude 3.5 Sonnet (October 2024) 56.9% native function calling Oct 2024 Official leaderboard
65 Palmyra X4 56.8% native function calling Oct 2024 Official leaderboard
66 Llama-3.1-Nemotron-Ultra-253B-v1 56.7% native function calling Apr 2025 Official leaderboard
67 Qwen3-1.7B 56.6% May 2025 Lab model card
68 Gemini 2.0 Flash-Lite 56.4% prompted Feb 2025 Official leaderboard
69 Amazon Nova Lite 55.9% native function calling Dec 2024 Official leaderboard
70 granite-4.0-1b 54.8% Oct 2025 Lab model card
71 Claude 3.5 Haiku 53.7% prompted Oct 2024 Official leaderboard
72 Grok-3 mini 53% native function calling Apr 2025 Official leaderboard
73 GPT-4.1 nano 53% native function calling Apr 2025 Official leaderboard
74 Llama 4 Maverick 52.9% native function calling Apr 2025 Official leaderboard
75 Llama 3.3 70B Instruct 52.2% native function calling Dec 2024 Official leaderboard
76 Amazon Nova Micro 51.8% native function calling Dec 2024 Official leaderboard
77 Command R7B 51.7% native function calling Dec 2024 Official leaderboard
78 Ministral 8B 51.5% native function calling Oct 2024 Official leaderboard
79 Mistral NeMo 51.5% native function calling Jul 2024 Official leaderboard
80 Gemma 3 12B 50.6% prompted Mar 2025 Official leaderboard
81 MiniCPM3-4B 50.3% native function calling Sep 2024 Official leaderboard
82 LFM2.5-1.2B-Instruct 49.1% Jan 2026 Lab model card
83 Granite-20B-FunctionCalling 48.9% native function calling Jul 2024 Official leaderboard
84 Command R+ 48% native function calling Aug 2024 Official leaderboard
85 Qwen3-0.6B 47.3% native function calling Apr 2025 Official leaderboard
86 Phi-4 47% prompted Dec 2024 Official leaderboard
87 Llama 3.2 3B 45.9% native function calling Sep 2024 Official leaderboard
88 Llama 4 Scout 45.4% native function calling Apr 2025 Official leaderboard
89 Falcon3-10B-Instruct 45.1% native function calling Dec 2024 Official leaderboard
90 Falcon3-7B-Instruct 43.6% native function calling Dec 2024 Official leaderboard
91 Gemma 3 4B 39.6% prompted Mar 2025 Official leaderboard
92 DBRX 39.2% prompted Mar 2024 Official leaderboard
93 GLM-4-9B-Chat 39.1% native function calling Jun 2024 Official leaderboard
94 Falcon3-3B-Instruct 32.5% native function calling Dec 2024 Official leaderboard
95 Llama 3.2 1B 21.9% native function calling Sep 2024 Official leaderboard
96 Falcon3-1B-Instruct 19.7% native function calling Dec 2024 Official leaderboard
97 Gemma 3 1B 16.3% prompted Mar 2025 Official leaderboard

The benchmark's own leaderboard has 89 entries, 67 of them models we track. Switch the scores above to rank them together.

Of runs with a cost, the best scores 71.7% for $0.008 a task

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

OpenAI Amazon Cohere Google DeepMind Mistral AI Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.0001 $0.001 $0.01 $0.10 $1.00 Cost per task, log scale GPT-4o GPT-4.1 mini GPT-4o mini Mistral Small 3.1 Amazon Nova Micro Command R7B 0% 20% 40% 60% 80% 100% $0.0001 $0.001 $0.01 $0.10 $1.00 Cost per task, log scale GPT-4o GPT-4.1 mini

Source: Official Berkeley Function Calling Leaderboard (BFCL) leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Llama-xLAM-2-70b-fc-r native function calling 78.5%
xLAM-2-32b-fc-r native function calling 76.4%
Llama-xLAM-2-8b-fc-r native function calling 72%
GPT-4o Nov 2024 native function calling 71.7% $0.008
GPT-4o Nov 2024 prompted 70.4% $0.01
GPT-4.5 native function calling 70.3% $0.24
Qwen3-32B native function calling 69.3%
GPT-4.1 native function calling 68.9% $0.007
GPT-4.1 prompted 68% $0.01
Qwen3-14B native function calling 68%
Qwen3-235B-A22B native function calling 67.9%
o1 prompted 67.8% $0.10
Qwen3-8B native function calling 66.3%
Qwen3-30B-A3B native function calling 65.7%
GPT-4.1 mini native function calling 65.2% $0.001
xLAM-2-3b-fc-r native function calling 64.8%
DeepSeek-V3 Mar 2025 native function calling 64.7%
o3-mini prompted 64.6% $0.008
Cohere Command A native function calling 64.4% $0.007
GPT-4o mini native function calling 64% $0.0005
DeepSeek-R1 May 2025 native function calling 63.8%
Qwen3-4B native function calling 63.4%
Grok 3 native function calling 63.3% $0.02
GPT-4.5 prompted 62.6% $0.47
GPT-4o mini prompted 61.8% $0.0009
Gemini 2.0 Flash Feb 2025 native function calling 61.2% $0.0006
Amazon Nova Pro native function calling 60.8% $0.005
GPT-4.1 mini prompted 60.6% $0.002
Gemini 2.0 Flash Feb 2025 prompted 60% $0.0008
Gemini 2.5 Pro May 2025 native function calling 59.9%
Gemma 3 27B prompted 59.1%
Gemini 2.0 Flash Thinking Jan 2025 prompted 58.7%
Claude 3.7 Sonnet native function calling 58.6% $0.003
QwQ-32B native function calling 58.3%
xLAM-2-1b-fc-r native function calling 58.2%
Mistral Large 2 Nov 2024 native function calling 58% $0.007
Mistral Small 3.1 native function calling 57.7% $0.0004
Claude 3 Opus native function calling 57% $0.02
Claude 3.5 Sonnet (October 2024) native function calling 56.9% $0.002
Palmyra X4 native function calling 56.8% $0.03
Llama-3.1-Nemotron-Ultra-253B-v1 native function calling 56.7%
Gemini 2.0 Flash-Lite prompted 56.4% $0.0003
Gemini 2.0 Flash-Lite native function calling 56.2% $0.0003
Grok 3 prompted 56.1% $0.02
o1 native function calling 56.1% $0.07
Amazon Nova Lite native function calling 55.9% $0.0004
Qwen3-1.7B native function calling 54.7%
Claude 3.5 Haiku prompted 53.7% $0.0005
Llama 3.1-70B prompted 53.6%
Grok-3 mini native function calling 53% $0.001
GPT-4.1 nano native function calling 53% $0.0004
Llama 4 Maverick native function calling 52.9%
Claude 3.5 Haiku native function calling 52.5% $0.0008
Llama 3.3 70B Instruct native function calling 52.2%
Amazon Nova Micro native function calling 51.8% $0.0002
Command R7B native function calling 51.7% $0.0001
Ministral 8B native function calling 51.5%
Mistral NeMo native function calling 51.5% $0.0006
o3-mini native function calling 51.3% $0.005
Gemma 3 12B prompted 50.6%
Claude 3.5 Sonnet (October 2024) prompted 50.5% $0.001
MiniCPM3-4B native function calling 50.3%
Claude 3.7 Sonnet prompted 50% $0.001
Claude 3 Opus prompted 49.7% $0.01
Llama 3.1-8B prompted 49.6%
GPT-4.1 nano prompted 49.4% $0.0006
Granite-20B-FunctionCalling native function calling 48.9%
Command R+ native function calling 48% $0.008
Qwen3-0.6B native function calling 47.3%
Phi-4 prompted 47%
Llama 3.2 3B native function calling 45.9%
Llama 4 Scout native function calling 45.4%
Falcon3-10B-Instruct native function calling 45.1%
Falcon3-7B-Instruct native function calling 43.6%
MiniCPM3-4B prompted 40.9%
Mistral NeMo prompted 39.7% $0.0009
Gemma 3 4B prompted 39.6%
DBRX prompted 39.2% $0.009
GLM-4-9B-Chat native function calling 39.1%
Falcon3-3B-Instruct native function calling 32.5%
Phi-4 Mini prompted 31.3%
Phi-4 Mini native function calling 28.2%
Llama 3.1-70B native function calling 27.1%
Llama 3.1-8B native function calling 25.9%
Llama 3.2 1B native function calling 21.9%
Falcon3-1B-Instruct native function calling 19.7%
Gemma 3 1B prompted 16.3%

2 models have held number one in labs’ own numbers since its release in February 2024 4 models have held number one across all independent runs since its release in February 2024 4 models have held number one with claims included since its release in February 2024

Lab claims count from the day the card came out, independent runs from the model’s release. We draw one row each for the benchmark’s own leaderboard and what labs claim in their cards.

Anthropic OpenAI Salesforce Z.ai Meta
Own leaderboard Replaced by BFCL V4, Jul 2025
Llama-xLAM-2-70b-fc-r 78.5%
Lab claims Passed 85% in Jul 2024, by Llama 3.1-405B
Llama 3.1-405B 88.5%

Sources: Official Berkeley Function Calling Leaderboard (BFCL) leaderboard and labs’ model cards

Meta’s own number for Llama 3.1-405B passed 85% in July 2024

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2024 2025 2026 2027 Near the ceiling (85%) Llama 3.1-405B GPT-4o mini ChatGLM3-6B 0% 20% 40% 60% 80% 100% 2024 2025 2026 2027 Near the ceiling (85%) Llama 3.1-405B GPT-4o mini

Sources: Official Berkeley Function Calling Leaderboard (BFCL) leaderboard and labs’ model cards

2 labs put it in a launch post in the last six months

Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.

Meta
Qwen
Reka
Cohere
NVIDIA
Z.ai
Amazon
DeepSeek
Tencent
Upstage
LG AI Research
Baidu
Meituan
IBM
Arcee AI
Liquid AI
MiniMax
Microsoft
DateLaunch postScoreAs the post writes it
Oct 23, 2026 Nemotron 3.5 Bolt E4B 67.3 BFCL-v3
Jun 2, 2026 MAI-Thinking-1 72 BFCL v3
May 28, 2026 LFM2.5-8B-A1B 64.4 BFCLv3
Apr 29, 2026 Granite 4.1 30B 73.7 BFCL v3
Mar 16, 2026 Nemotron 3 Nano 4B 61.1 BFCL v3
Feb 12, 2026 MiniMax-M2.5 BFCL multi-turn
Jan 20, 2026 LFM2.5-1.2B-Thinking 57 BFCLv3
Jan 5, 2026 LFM2.5-1.2B-Instruct 49.1 BFCLv3
Dec 1, 2025 Trinity Mini 59.7 BFCL V3
Oct 28, 2025 granite-4.0-1b 54.8 BFCL v3
Oct 2, 2025 Granite 4.0 H Small 64.7 BFCL v3
Sep 23, 2025 Qwen3-VL-235B-A22B 71.9 BFCL
Sep 22, 2025 Qwen3-Omni-30B-A3B 64.4 BFCL-v3
Sep 21, 2025 LongCat-Flash-Thinking 74.4 BFCL V3(full)
Sep 11, 2025 Qwen3-Next-80B-A3B 72 BFCL-v3
Sep 9, 2025 ERNIE-4.5-21B-A3B-Thinking 65 BFCL (Accuracy)
Aug 21, 2025 Command A Reasoning 70.3 BFCL-v3 (FC setting)
Aug 18, 2025 nvidia-nemotron-nano-9b-v2 66.9 BFCL v3
Jul 28, 2025 GLM-4.5 77.8 BFCL v3 (Full)
Jul 25, 2025 Llama-3.3-Nemotron-Super-49B-v1.5 71.8 BFCL v3
Jul 25, 2025 Qwen3-235B-A22B-Thinking 71.9 BFCL-v3
Jul 22, 2025 Qwen3-Coder-480B-A35B 68.7 BFCL-v3
Jul 15, 2025 EXAONE 4.0 32B 65.2 BFCL-v3 65.2 non-reasoning (63.9 reasoning)
Jul 10, 2025 Solar Pro 2 63.9 BFCL.v3
Jun 27, 2025 Hunyuan-A13B-Instruct 78.3 BFCL v3
May 28, 2025 DeepSeek-R1 BFCL_v3_MultiTurn (Acc)
Apr 30, 2025 Amazon Nova Premier 63.7 BFCL (2025-04-25)
Apr 29, 2025 Qwen3-235B-A22B 70.8 BFCL v3
Apr 14, 2025 GLM-Z1-32B-0414 70.5 BFCL (v3)
Apr 7, 2025 Llama-3.1-Nemotron-Ultra-253B-v1 BFCL V2 Live (reasoning on)
Mar 18, 2025 Llama-3.3-Nemotron-Super-49B-v1 BFCL V2 Live (reasoning off)
Mar 13, 2025 Cohere Command A BFCL (BFCL-v3, 2025-03-12, chart only)
Mar 10, 2025 Reka Flash 3 BFCL v2
Mar 6, 2025 QwQ-32B BFCL
Dec 6, 2024 Llama 3.3 70B Instruct BFCL v2, overall_ast_summary/macro_avg/valid
Jul 23, 2024 Llama 3.1-405B 88.5 BFCL

31 system cards report it, from 11 labs

Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.

DateDocumentScoreAs the document writes itSection
Oct 23, 2026 Nemotron 3.5 Bolt E4B technical report 67.3 BFCL-v3, released Q4_K_M imatrix checkpoint (Table 5). BF16 68.51 (Table 3) Capabilities
Jun 2, 2026 MAI-Thinking-1 technical report 72 BFCL v3 (Table 12, rounded) Capabilities
Nov 26, 2025 Qwen3-VL-235B-A22B technical report 71.8 BFCL-v3 (Table 6) Capabilities
Oct 21, 2025 Ring-1T technical report 68.8 BFCL v3 (Table 1) Capabilities
Sep 23, 2025 LongCat-Flash-Thinking technical report 74.4 BFCL V3(full) (Table 2) Capabilities
Sep 22, 2025 Qwen3-Omni-30B-A3B technical report 64.4 BFCL-v3, Instruct (Table 4). Thinking 63.2 (Table 5) Capabilities
Aug 20, 2025 nvidia-nemotron-nano-9b-v2 technical report 66.9 BFCLv3, NVIDIA-Nemotron-Nano-9B-v2 (Figure 1) Capabilities
Aug 8, 2025 GLM-4.5 technical report 77.8 BFCL V3 (Table 3) Capabilities
Aug 8, 2025 GLM-4.5-Air technical report 76.4 BFCL V3 (Table 3) Capabilities
Jul 15, 2025 EXAONE 4.0 32B technical report 63.9 BFCL-v3 (Table 3, Reasoning). Non-reasoning 65.2 (Table 4) Capabilities
Jun 26, 2025 Hunyuan-A13B-Instruct technical report 78.3 BFCL v3, slow-thinking (Table 3). Fast-thinking 65.9 (Table 4) Capabilities
May 14, 2025 Qwen3-8B technical report 68.1 BFCL v3, thinking mode (Table 17). Non-thinking 60.2 (Table 18) Capabilities
May 14, 2025 Qwen3-4B technical report 65.9 BFCL v3, thinking mode (Table 17). Non-thinking 57.6 (Table 18) Capabilities
May 14, 2025 Qwen3-32B technical report 70.3 BFCL v3, thinking mode (Table 13). Non-thinking 63.0 (Table 14) Capabilities
May 14, 2025 Qwen3-30B-A3B technical report 69.1 BFCL v3, thinking mode (Table 15). Non-thinking 58.6 (Table 16) Capabilities
May 14, 2025 Qwen3-235B-A22B technical report 70.8 BFCL v3, thinking mode (Table 11). Non-thinking 68.0 (Table 12) Capabilities
May 14, 2025 Qwen3-14B technical report 70.4 BFCL v3, thinking mode (Table 15). Non-thinking 61.5 (Table 16) Capabilities
May 14, 2025 Qwen3-1.7B technical report 56.6 BFCL v3, thinking mode (Table 19). Non-thinking 52.2 (Table 20) Capabilities
May 2, 2025 Llama-3.3-Nemotron-Super-49B-v1 technical report BFCL V2 Live, reasoning on, reasoning off 73.9 (Table 4) Capabilities
May 2, 2025 Llama-3.1-Nemotron-Ultra-253B-v1 technical report BFCL V2 Live, reasoning on, reasoning off 73.6 (Table 5) Capabilities
Apr 30, 2025 Amazon Nova Premier technical report 63.7 BFCL (2025-04-25), BFCL V3 overall accuracy (Table 1) Capabilities
Apr 1, 2025 Cohere Command A technical report 63.8 BFCL Overall from the official leaderboard (Table 5). Live AST 80.5, Multi-turn 25.5 Capabilities
Mar 3, 2025 Phi-4 Mini technical report 70.3 BFCL (0-shot), version not stated (Table 7) Capabilities
Dec 6, 2024 Llama 3.3 70B Instruct model card BFCL v2, 0-shot, overall_ast_summary/macro_avg/valid (Instruction tuned models table) Capabilities
Nov 23, 2024 Llama 3.1-8B technical report 76.1 BFCL, post-trained (Table 2, Table 22) Capabilities
Nov 23, 2024 Llama 3.1-70B technical report 84.8 BFCL, post-trained (Table 2, Table 22) Capabilities
Nov 23, 2024 Llama 3.1-405B technical report 88.5 BFCL, post-trained (Table 2, Table 22) Capabilities
Jul 23, 2024 Llama 3.1-8B model card 76.1 BFCL, 0-shot, acc, instruction tuned (Instruction tuned models table) Capabilities
Jul 23, 2024 Llama 3.1-70B model card 84.8 BFCL, 0-shot, acc, instruction tuned (Instruction tuned models table) Capabilities
Jul 23, 2024 Llama 3.1-405B model card 88.5 BFCL, 0-shot, acc, instruction tuned (Instruction tuned models table) Capabilities
Jun 18, 2024 ChatGLM3-6B technical report 57.9 ChatGLM3-6B, Berkeley Function Call Leaderboard Overall (Table 7) Capabilities

The lab's number comes out ahead in 10 of 11 pairs that differ

Each score a lab reported for its own model, beside the same model's best score on the leaderboard. Gaps mostly come from different harnesses, effort levels and task subsets. Every benchmark.

Model Lab says Board says Gap
Phi-4 Mini 70.3% BFCL (0-shot), version not stated (Table 7) 31.3% Leaderboard, prompted +39 pts
Llama 3.1-70B 84.8% BFCL, 0-shot, acc, instruction tuned (Instruction tuned models table) 53.6% Leaderboard, prompted +31 pts
Llama 3.1-8B 76.1% BFCL, 0-shot, acc, instruction tuned (Instruction tuned models table) 49.6% Leaderboard, prompted +27 pts
Qwen3-30B-A3B 69.1% BFCL v3, thinking mode (Table 15). Non-thinking 58.6 (Table 16) 65.7% Leaderboard, native function calling +3.4 pts
Qwen3-235B-A22B 70.8% BFCL v3 67.9% Leaderboard, native function calling +2.9 pts
Qwen3-4B 65.9% BFCL v3, thinking mode (Table 17). Non-thinking 57.6 (Table 18) 63.4% Leaderboard, native function calling +2.5 pts
Qwen3-14B 70.4% BFCL v3, thinking mode (Table 15). Non-thinking 61.5 (Table 16) 68% Leaderboard, native function calling +2.4 pts
Qwen3-1.7B 56.6% BFCL v3, thinking mode (Table 19). Non-thinking 52.2 (Table 20) 54.7% Leaderboard, native function calling +1.9 pts
Qwen3-8B 68.1% BFCL v3, thinking mode (Table 17). Non-thinking 60.2 (Table 18) 66.3% Leaderboard, native function calling +1.8 pts
Qwen3-32B 70.3% BFCL v3, thinking mode (Table 13). Non-thinking 63.0 (Table 14) 69.3% Leaderboard, native function calling +1.0 pts
Cohere Command A 63.8% BFCL Overall from the official leaderboard (Table 5). Live AST 80.5, Multi-turn 25.5 64.4% Leaderboard, native function calling −0.6 pts

5 logged changes that move scores

Changes by the maintainers, so a score from before a change may not compare with one from after it. Newest first, each with the maintainers' own source. The list may be incomplete. What we log.

  1. Grader changed October 30th, 2024

    The maintainers required multi-turn answers to pass both response and state checks. Earlier evaluations checked state alone. Source

  2. Questions corrected October 16th, 2024

    The maintainers corrected questions and reference answers across six categories. Earlier evaluations used different task definitions. Source

  3. Tasks changed September 19th, 2024

    The maintainers added multi-turn tasks and state-based evaluation in BFCL V3. Its coverage differs from earlier versions. Source

  4. Tasks changed August 19th, 2024

    The maintainers added user-contributed live tasks in BFCL V2. Its coverage differs from V1. Source

  5. Questions corrected July 6th, 2024

    The maintainers corrected reference answers for 39 parallel-call tasks. Earlier scores used different answer keys. Source

Versions

We switch the top row to each new version when it comes out. Open it for one row per version. An old version stays in play while people still test it, so whoever first takes it past 85% gets credit for closing it out.

Z.ai Meta Amazon Qwen ByteDance Seed
BFCL V4 6 leaders
Qwen3.7-Max 75%
Berkeley Function Calling Leaderboard (BFCL) Passed 85% in Jul 2024, by Llama 3.1-405B
Llama 3.1-405B 88.5%
BFCL V4 4 leaders
Qwen3.7-Max 75%

Source: Labs’ model cards

Replaced by BFCL V4

More in Tool use

All 144 in Tool use
  • AutomationBench In use Agent completes end-to-end business workflows inside a simulated company exposing dozens of REST APIs. Graded on resulting state. Gemini 4 Carbon 53.9%
  • MCP Atlas In use Multi-step workflows over real MCP servers. Scores tool-use competency on realistic tasks. Gemini 4 Argon 90.7%
  • τ³-bench Banking In use Agents act as banking customer-service reps over many turns with a simulated user, following policy. Pass^1 on the final database state. Qwen 3.8 Max 55.1%
  • Toolathlon-Verified In use 108 long-horizon tasks across 32 apps through about 600 MCP tools. Execution-based checkers inspect final environment state. GLM-5.3-Flash 78.4%
  • DeepSearchQA In use Questions that need multi-step search and synthesis. Final answers graded against references. Kimi K3 95%
  • Claw-Eval In use 300 human-verified agent tasks (service orchestration, multimodal, multi-turn dialogue) audited via traces and snapshots. Rubric scores with Pass^3 consistency. Claude Sonnet 4.6 81.4%