Benchmark Almanac

Berkeley Function Calling Leaderboard (BFCL)

Llama 3.1-405B 88.5% leads since Jul 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.

Made by UC Berkeley (Gorilla) Released Feb 2024 Function calling All tasks public Last in a launch post Oct 2026 Last model added Oct 2026 Scores refreshed 2 Nov 2026
Level
Everyday Most adults, without training
Scored
Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
Answers
Agent task Many steps in a terminal, browser, desktop, app, game or robot.
ShishirPatil/gorilla 13,062 stars Leaderboard
47
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.

47 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. 77.3 Llama 3.3 70B Instruct
  6. 77.2 Reka Flash 3
  7. 76.8 MiniMax-M2.5
  8. 76.4 GLM-4.5-Air
  9. 76.1 Llama 3.1-8B
  10. 74.4 LongCat-Flash-Thinking
  11. 74.1 Llama-3.1-Nemotron-Ultra-253B-v1
  12. 73.7 Llama-3.3-Nemotron-Super-49B-v1
  13. 73.7 Granite 4.1 30B
  14. 72.0 Qwen3-Next-80B-A3B
  15. 72.0 MAI-Thinking-1
  16. 71.9 Qwen3-235B-A22B-Thinking
  17. 71.9 Qwen3-VL-235B-A22B
  18. 71.8 Llama-3.3-Nemotron-Super-49B-v1.5
  19. 70.8 Qwen3-235B-A22B
  20. 70.5 GLM-Z1-32B-0414
  21. 70.4 Qwen3-14B
  22. 70.3 Phi-4 Mini
  23. 70.3 Qwen3-32B
  24. 70.3 Command A Reasoning
  25. 69.1 Qwen3-30B-A3B
  26. 68.8 Ring-1T
  27. 68.7 Qwen3-Coder-480B-A35B
  28. 68.1 Qwen3-8B
  29. 67.3 Nemotron 3.5 Bolt E4B
  30. 66.9 nvidia-nemotron-nano-9b-v2
  31. 65.9 Qwen3-4B
  32. 65.2 EXAONE 4.0 32B
  33. 65.0 ERNIE-4.5-21B-A3B-Thinking
  34. 64.7 Granite 4.0 H Small
  35. 64.4 Qwen3-Omni-30B-A3B
  36. 64.4 LFM2.5-8B-A1B
  37. 63.9 Solar Pro 2
  38. 63.8 Cohere Command A
  39. 63.7 Amazon Nova Premier
  40. 61.1 Nemotron 3 Nano 4B
  41. 59.7 Trinity Mini
  42. 57.9 ChatGLM3-6B
  43. 57.0 LFM2.5-1.2B-Thinking
  44. 56.6 Qwen3-1.7B
  45. 54.8 granite-4.0-1b
  46. 49.1 LFM2.5-1.2B-Instruct
  47. 37.0 DeepSeek-R1
  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 Llama 3.3 70B Instruct 77.3%
  6. 6 Reka Flash 3 77.2%
  7. 7 MiniMax-M2.5 76.8%
  8. 8 GLM-4.5-Air 76.4%
  9. 9 Llama 3.1-8B 76.1%
  10. 10 LongCat-Flash-Thinking 74.4%
  11. 11 Llama-3.1-Nemotron-Ultra-253B-v1 74.1%
  12. 12 Llama-3.3-Nemotron-Super-49B-v1 73.7%
  13. 13 Granite 4.1 30B 73.7%
  14. 14 Qwen3-Next-80B-A3B 72%
  15. 15 MAI-Thinking-1 72%
  16. 16 Qwen3-235B-A22B-Thinking Jul 2025 71.9%
  17. 17 Qwen3-VL-235B-A22B 71.9%
  18. 18 Llama-3.3-Nemotron-Super-49B-v1.5 71.8%
  19. 19 Qwen3-235B-A22B 70.8%
  20. 20 GLM-Z1-32B-0414 70.5%
  21. 21 Qwen3-14B 70.4%
  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 Ring-1T 68.8%
  27. 27 Qwen3-Coder-480B-A35B 68.7%
  28. 28 Qwen3-8B 68.1%
  29. 29 Nemotron 3.5 Bolt E4B 67.3%
  30. 30 nvidia-nemotron-nano-9b-v2 66.9%
  31. 31 Qwen3-4B 65.9%
  32. 32 EXAONE 4.0 32B 65.2%
  33. 33 ERNIE-4.5-21B-A3B-Thinking 65%
  34. 34 Granite 4.0 H Small 64.7%
  35. 35 Qwen3-Omni-30B-A3B 64.4%
  36. 36 LFM2.5-8B-A1B 64.4%
  37. 37 Solar Pro 2 63.9%
  38. 38 Cohere Command A 63.8%
  39. 39 Amazon Nova Premier 63.7%
  40. 40 Nemotron 3 Nano 4B 61.1%
  41. 41 Trinity Mini 59.7%
  42. 42 ChatGLM3-6B 57.9%
  43. 43 LFM2.5-1.2B-Thinking 57%
  44. 44 Qwen3-1.7B 56.6%
  45. 45 granite-4.0-1b 54.8%
  46. 46 LFM2.5-1.2B-Instruct 49.1%
  47. 47 DeepSeek-R1 37%
#ModelScoreSettingDateSource
1 Llama 3.1-405B 88.5% Jul 2024 Model card
2 Llama 3.1-70B 84.8% Jul 2024 Model card
3 Hunyuan-A13B-Instruct 78.3% Jun 2025 Model card
4 GLM-4.5 77.8% Jul 2025 Model card
5 Llama 3.3 70B Instruct 77.3% Dec 2024 Model card
6 Reka Flash 3 77.2% Mar 2025 Model card
7 MiniMax-M2.5 76.8% Feb 2026 Model card
8 GLM-4.5-Air 76.4% Aug 2025 Model card
9 Llama 3.1-8B 76.1% Jul 2024 Model card
10 LongCat-Flash-Thinking 74.4% Sep 2025 Model card
11 Llama-3.1-Nemotron-Ultra-253B-v1 74.1% Apr 2025 Model card
12 Llama-3.3-Nemotron-Super-49B-v1 73.7% Mar 2025 Model card
13 Granite 4.1 30B 73.7% Apr 2026 Model card
14 Qwen3-Next-80B-A3B 72% Sep 2025 Model card
15 MAI-Thinking-1 72% Jun 2026 Model card
16 Qwen3-235B-A22B-Thinking Jul 2025 71.9% Jul 2025 Model card
17 Qwen3-VL-235B-A22B 71.9% Sep 2025 Model card
18 Llama-3.3-Nemotron-Super-49B-v1.5 71.8% Jul 2025 Model card
19 Qwen3-235B-A22B 70.8% Apr 2025 Model card
20 GLM-Z1-32B-0414 70.5% Apr 2025 Model card
21 Qwen3-14B 70.4% May 2025 Model card
22 Phi-4 Mini 70.3% Mar 2025 Model card
23 Qwen3-32B 70.3% May 2025 Model card
24 Command A Reasoning 70.3% Aug 2025 Model card
25 Qwen3-30B-A3B 69.1% May 2025 Model card
26 Ring-1T 68.8% Oct 2025 Model card
27 Qwen3-Coder-480B-A35B 68.7% Jul 2025 Model card
28 Qwen3-8B 68.1% May 2025 Model card
29 Nemotron 3.5 Bolt E4B 67.3% Oct 2026 Model card
30 nvidia-nemotron-nano-9b-v2 66.9% Aug 2025 Model card
31 Qwen3-4B 65.9% May 2025 Model card
32 EXAONE 4.0 32B 65.2% Jul 2025 Model card
33 ERNIE-4.5-21B-A3B-Thinking 65% Sep 2025 Model card
34 Granite 4.0 H Small 64.7% Oct 2025 Model card
35 Qwen3-Omni-30B-A3B 64.4% Sep 2025 Model card
36 LFM2.5-8B-A1B 64.4% May 2026 Model card
37 Solar Pro 2 63.9% Jul 2025 Model card
38 Cohere Command A 63.8% Apr 2025 Model card
39 Amazon Nova Premier 63.7% Apr 2025 Model card
40 Nemotron 3 Nano 4B 61.1% Mar 2026 Model card
41 Trinity Mini 59.7% Dec 2025 Model card
42 ChatGLM3-6B 57.9% Jun 2024 Model card
43 LFM2.5-1.2B-Thinking 57% Jan 2026 Model card
44 Qwen3-1.7B 56.6% May 2025 Model card
45 granite-4.0-1b 54.8% Oct 2025 Model card
46 LFM2.5-1.2B-Instruct 49.1% Jan 2026 Model card
47 DeepSeek-R1 37% May 2025 Model card

The benchmark's own leaderboard has 89 entries, 67 of them models we track. We plot every source in the charts below.

4 models have held number one since its release in Feb 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%

The best score passed 85% in Jul 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

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
23 Oct 2026 Nemotron 3.5 Bolt E4B 67.3 BFCL-v3
2 Jun 2026 MAI-Thinking-1 72 BFCL v3
28 May 2026 LFM2.5-8B-A1B 64.4 BFCLv3
29 Apr 2026 Granite 4.1 30B 73.7 BFCL v3
16 Mar 2026 Nemotron 3 Nano 4B 61.1 BFCL v3
12 Feb 2026 MiniMax-M2.5 76.8 BFCL multi-turn
20 Jan 2026 LFM2.5-1.2B-Thinking 57 BFCLv3
5 Jan 2026 LFM2.5-1.2B-Instruct 49.1 BFCLv3
1 Dec 2025 Trinity Mini 59.7 BFCL V3
28 Oct 2025 granite-4.0-1b 54.8 BFCL v3
2 Oct 2025 Granite 4.0 H Small 64.7 BFCL v3
23 Sep 2025 Qwen3-VL-235B-A22B 71.9 BFCL
22 Sep 2025 Qwen3-Omni-30B-A3B 64.4 BFCL-v3
21 Sep 2025 LongCat-Flash-Thinking 74.4 BFCL V3(full)
11 Sep 2025 Qwen3-Next-80B-A3B 72 BFCL-v3
9 Sep 2025 ERNIE-4.5-21B-A3B-Thinking 65 BFCL (Accuracy)
21 Aug 2025 Command A Reasoning 70.3 BFCL-v3 (FC setting)
18 Aug 2025 nvidia-nemotron-nano-9b-v2 66.9 BFCL v3
28 Jul 2025 GLM-4.5 77.8 BFCL v3 (Full)
25 Jul 2025 Llama-3.3-Nemotron-Super-49B-v1.5 71.8 BFCL v3
25 Jul 2025 Qwen3-235B-A22B-Thinking 71.9 BFCL-v3
22 Jul 2025 Qwen3-Coder-480B-A35B 68.7 BFCL-v3
15 Jul 2025 EXAONE 4.0 32B 65.2 BFCL-v3 65.2 non-reasoning (63.9 reasoning)
10 Jul 2025 Solar Pro 2 63.9 BFCL.v3
27 Jun 2025 Hunyuan-A13B-Instruct 78.3 BFCL v3
28 May 2025 DeepSeek-R1 37 BFCL_v3_MultiTurn (Acc)
30 Apr 2025 Amazon Nova Premier 63.7 BFCL (2025-04-25)
29 Apr 2025 Qwen3-235B-A22B 70.8 BFCL v3
14 Apr 2025 GLM-Z1-32B-0414 70.5 BFCL (v3)
7 Apr 2025 Llama-3.1-Nemotron-Ultra-253B-v1 74.1 BFCL V2 Live (reasoning on)
18 Mar 2025 Llama-3.3-Nemotron-Super-49B-v1 73.7 BFCL V2 Live (reasoning off)
13 Mar 2025 Cohere Command A BFCL (BFCL-v3, 2025-03-12, chart only)
10 Mar 2025 Reka Flash 3 77.2 BFCL v2
6 Mar 2025 QwQ-32B BFCL
6 Dec 2024 Llama 3.3 70B Instruct 77.3 BFCL v2, overall_ast_summary/macro_avg/valid
23 Jul 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
23 Oct 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
2 Jun 2026 MAI-Thinking-1 technical report 72 BFCL v3 (Table 12, rounded) Capabilities
26 Nov 2025 Qwen3-VL-235B-A22B technical report 71.8 BFCL-v3 (Table 6) Capabilities
21 Oct 2025 Ring-1T technical report 68.8 BFCL v3 (Table 1) Capabilities
23 Sep 2025 LongCat-Flash-Thinking technical report 74.4 BFCL V3(full) (Table 2) Capabilities
22 Sep 2025 Qwen3-Omni-30B-A3B technical report 64.4 BFCL-v3, Instruct (Table 4). Thinking 63.2 (Table 5) Capabilities
20 Aug 2025 nvidia-nemotron-nano-9b-v2 technical report 66.9 BFCLv3, NVIDIA-Nemotron-Nano-9B-v2 (Figure 1) Capabilities
8 Aug 2025 GLM-4.5 technical report 77.8 BFCL V3 (Table 3) Capabilities
8 Aug 2025 GLM-4.5-Air technical report 76.4 BFCL V3 (Table 3) Capabilities
15 Jul 2025 EXAONE 4.0 32B technical report 63.9 BFCL-v3 (Table 3, Reasoning). Non-reasoning 65.2 (Table 4) Capabilities
26 Jun 2025 Hunyuan-A13B-Instruct technical report 78.3 BFCL v3, slow-thinking (Table 3). Fast-thinking 65.9 (Table 4) Capabilities
14 May 2025 Qwen3-8B technical report 68.1 BFCL v3, thinking mode (Table 17). Non-thinking 60.2 (Table 18) Capabilities
14 May 2025 Qwen3-4B technical report 65.9 BFCL v3, thinking mode (Table 17). Non-thinking 57.6 (Table 18) Capabilities
14 May 2025 Qwen3-32B technical report 70.3 BFCL v3, thinking mode (Table 13). Non-thinking 63.0 (Table 14) Capabilities
14 May 2025 Qwen3-30B-A3B technical report 69.1 BFCL v3, thinking mode (Table 15). Non-thinking 58.6 (Table 16) Capabilities
14 May 2025 Qwen3-235B-A22B technical report 70.8 BFCL v3, thinking mode (Table 11). Non-thinking 68.0 (Table 12) Capabilities
14 May 2025 Qwen3-14B technical report 70.4 BFCL v3, thinking mode (Table 15). Non-thinking 61.5 (Table 16) Capabilities
14 May 2025 Qwen3-1.7B technical report 56.6 BFCL v3, thinking mode (Table 19). Non-thinking 52.2 (Table 20) Capabilities
2 May 2025 Llama-3.3-Nemotron-Super-49B-v1 technical report 73.7 BFCL V2 Live, reasoning on, reasoning off 73.9 (Table 4) Capabilities
2 May 2025 Llama-3.1-Nemotron-Ultra-253B-v1 technical report 74.1 BFCL V2 Live, reasoning on, reasoning off 73.6 (Table 5) Capabilities
30 Apr 2025 Amazon Nova Premier technical report 63.7 BFCL (2025-04-25), BFCL V3 overall accuracy (Table 1) Capabilities
1 Apr 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
3 Mar 2025 Phi-4 Mini technical report 70.3 BFCL (0-shot), version not stated (Table 7) Capabilities
6 Dec 2024 Llama 3.3 70B Instruct model card 77.3 BFCL v2, 0-shot, overall_ast_summary/macro_avg/valid (Instruction tuned models table) Capabilities
23 Nov 2024 Llama 3.1-8B technical report 76.1 BFCL, post-trained (Table 2, Table 22) Capabilities
23 Nov 2024 Llama 3.1-70B technical report 84.8 BFCL, post-trained (Table 2, Table 22) Capabilities
23 Nov 2024 Llama 3.1-405B technical report 88.5 BFCL, post-trained (Table 2, Table 22) Capabilities
23 Jul 2024 Llama 3.1-8B model card 76.1 BFCL, 0-shot, acc, instruction tuned (Instruction tuned models table) Capabilities
23 Jul 2024 Llama 3.1-70B model card 84.8 BFCL, 0-shot, acc, instruction tuned (Instruction tuned models table) Capabilities
23 Jul 2024 Llama 3.1-405B model card 88.5 BFCL, 0-shot, acc, instruction tuned (Instruction tuned models table) Capabilities
18 Jun 2024 ChatGLM3-6B technical report 57.9 ChatGLM3-6B, Berkeley Function Call Leaderboard Overall (Table 7) Capabilities

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.

Anthropic Z.ai OpenAI Meta
BFCL V4 7 leaders
Claude Opus 4.5 77.5%
Berkeley Function Calling Leaderboard (BFCL) Passed 85% in Jul 2024, by Llama 3.1-405B
Llama 3.1-405B 88.5%
BFCL V4 3 leaders
Claude Opus 4.5 77.5%
Replaced by BFCL V4

More in Tool use

All 125 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. Muse Spark 1.1 88.1%
  • 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%
  • 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%
  • DeepSearchQA In use Questions that need multi-step search and synthesis. Final answers graded against references. Kimi K3 95%
  • MCPMark In use Agent stress-tests realistic MCP tool use (GitHub, Notion, Postgres, browser). Graded by programmatic verification scripts. Kimi K3 94.5%