Datasets:
id stringlengths 35 88 | modelVariant stringlengths 7 49 | modelVersion null | aiSystem null | aiCompany stringclasses 11
values | aiModality stringclasses 2
values | failureType stringclasses 2
values | metricName stringclasses 7
values | metricValue float64 0 1.5k | metricUnit stringclasses 2
values | metricDirection stringclasses 2
values | rank float64 1 109 ⌀ | source stringclasses 3
values | sourceUrl stringclasses 3
values | sourceLicense stringclasses 2
values | methodology stringclasses 3
values | measuredAt stringclasses 7
values | fetchedAt stringclasses 18
values |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
bfcl-amazon-nova-2-lite-v1-0-fc-function_calling_accuracy-2026-05-17 | Amazon-Nova-2-Lite-v1:0 (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 27.1 | percentage | higher_is_better | 80 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-amazon-nova-micro-v1-0-fc-function_calling_accuracy-2026-05-17 | Amazon-Nova-Micro-v1:0 (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 22.29 | percentage | higher_is_better | 95 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-amazon-nova-pro-v1-0-fc-function_calling_accuracy-2026-05-17 | Amazon-Nova-Pro-v1:0 (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 24.97 | percentage | higher_is_better | 88 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-arch-agent-1-5b-function_calling_accuracy-2026-05-17 | Arch-Agent-1.5B | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 32.14 | percentage | higher_is_better | 60 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
bfcl-arch-agent-32b-function_calling_accuracy-2026-05-17 | Arch-Agent-32B | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 45.37 | percentage | higher_is_better | 37 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.216Z |
bfcl-arch-agent-3b-function_calling_accuracy-2026-05-17 | Arch-Agent-3B | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 35.36 | percentage | higher_is_better | 56 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
bfcl-bielik-11b-v2-3-instruct-prompt-function_calling_accuracy-2026-05-17 | Bielik-11B-v2.3-Instruct (Prompt) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 21.9 | percentage | higher_is_better | 99 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.219Z |
bfcl-bitagent-bounty-8b-function_calling_accuracy-2026-05-17 | BitAgent-Bounty-8B | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 46.23 | percentage | higher_is_better | 36 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.216Z |
bfcl-claude-haiku-4-5-20251001-fc-function_calling_accuracy-2026-05-17 | Claude-Haiku-4-5-20251001 (FC) | null | null | Anthropic | agente autónomo | fallo técnico | function_calling_accuracy | 68.7 | percentage | higher_is_better | 6 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.213Z |
bfcl-claude-haiku-4-5-20251001-prompt-function_calling_accuracy-2026-05-17 | Claude-Haiku-4-5-20251001 (Prompt) | null | null | Anthropic | agente autónomo | fallo técnico | function_calling_accuracy | 25.26 | percentage | higher_is_better | 87 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-claude-opus-4-5-20251101-fc-function_calling_accuracy-2026-05-17 | Claude-Opus-4-5-20251101 (FC) | null | null | Anthropic | agente autónomo | fallo técnico | function_calling_accuracy | 77.47 | percentage | higher_is_better | 1 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.213Z |
bfcl-claude-opus-4-5-20251101-prompt-function_calling_accuracy-2026-05-17 | Claude-Opus-4-5-20251101 (Prompt) | null | null | Anthropic | agente autónomo | fallo técnico | function_calling_accuracy | 33.47 | percentage | higher_is_better | 57 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
bfcl-claude-sonnet-4-5-20250929-fc-function_calling_accuracy-2026-05-17 | Claude-Sonnet-4-5-20250929 (FC) | null | null | Anthropic | agente autónomo | fallo técnico | function_calling_accuracy | 73.24 | percentage | higher_is_better | 2 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.213Z |
bfcl-claude-sonnet-4-5-20250929-prompt-function_calling_accuracy-2026-05-17 | Claude-Sonnet-4-5-20250929 (Prompt) | null | null | Anthropic | agente autónomo | fallo técnico | function_calling_accuracy | 24.9 | percentage | higher_is_better | 89 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-coalm-70b-function_calling_accuracy-2026-05-17 | CoALM-70B | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 27.99 | percentage | higher_is_better | 74 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-coalm-8b-function_calling_accuracy-2026-05-17 | CoALM-8B | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 26.81 | percentage | higher_is_better | 84 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-command-a-fc-function_calling_accuracy-2026-05-17 | Command A (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 46.49 | percentage | higher_is_better | 35 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.216Z |
bfcl-command-a-reasoning-fc-function_calling_accuracy-2026-05-17 | Command A Reasoning (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 57.06 | percentage | higher_is_better | 13 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.214Z |
bfcl-command-r7b-fc-function_calling_accuracy-2026-05-17 | Command R7B (FC) | null | null | Cohere | agente autónomo | fallo técnico | function_calling_accuracy | 32.07 | percentage | higher_is_better | 61 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
bfcl-deepseek-v3-2-exp-fc-function_calling_accuracy-2026-05-17 | DeepSeek-V3.2-Exp (FC) | null | null | DeepSeek | agente autónomo | fallo técnico | function_calling_accuracy | 54.12 | percentage | higher_is_better | 19 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.214Z |
bfcl-deepseek-v3-2-exp-prompt-thinking-function_calling_accuracy-2026-05-17 | DeepSeek-V3.2-Exp (Prompt + Thinking) | null | null | DeepSeek | agente autónomo | fallo técnico | function_calling_accuracy | 56.73 | percentage | higher_is_better | 14 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.214Z |
bfcl-falcon3-10b-instruct-fc-function_calling_accuracy-2026-05-17 | Falcon3-10B-Instruct (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 27.01 | percentage | higher_is_better | 82 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-falcon3-1b-instruct-fc-function_calling_accuracy-2026-05-17 | Falcon3-1B-Instruct (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 11.08 | percentage | higher_is_better | 106 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.219Z |
bfcl-falcon3-3b-instruct-fc-function_calling_accuracy-2026-05-17 | Falcon3-3B-Instruct (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 16.25 | percentage | higher_is_better | 104 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.219Z |
bfcl-falcon3-7b-instruct-fc-function_calling_accuracy-2026-05-17 | Falcon3-7B-Instruct (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 24.03 | percentage | higher_is_better | 91 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-gemini-2-5-flash-fc-function_calling_accuracy-2026-05-17 | Gemini-2.5-Flash (FC) | null | null | Google | agente autónomo | fallo técnico | function_calling_accuracy | 56.24 | percentage | higher_is_better | 15 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.214Z |
bfcl-gemini-2-5-flash-lite-fc-function_calling_accuracy-2026-05-17 | Gemini-2.5-Flash-Lite (FC) | null | null | Google | agente autónomo | fallo técnico | function_calling_accuracy | 36.87 | percentage | higher_is_better | 52 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
bfcl-gemini-2-5-flash-lite-prompt-function_calling_accuracy-2026-05-17 | Gemini-2.5-Flash-Lite (Prompt) | null | null | Google | agente autónomo | fallo técnico | function_calling_accuracy | 28.03 | percentage | higher_is_better | 73 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-gemini-2-5-flash-prompt-function_calling_accuracy-2026-05-17 | Gemini-2.5-Flash (Prompt) | null | null | Google | agente autónomo | fallo técnico | function_calling_accuracy | 50.9 | percentage | higher_is_better | 26 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.215Z |
bfcl-gemini-3-pro-preview-fc-function_calling_accuracy-2026-05-17 | Gemini-3-Pro-Preview (FC) | null | null | Google | agente autónomo | fallo técnico | function_calling_accuracy | 68.14 | percentage | higher_is_better | 7 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.213Z |
bfcl-gemini-3-pro-preview-prompt-function_calling_accuracy-2026-05-17 | Gemini-3-Pro-Preview (Prompt) | null | null | Google | agente autónomo | fallo técnico | function_calling_accuracy | 72.51 | percentage | higher_is_better | 3 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.213Z |
bfcl-gemma-3-12b-it-prompt-function_calling_accuracy-2026-05-17 | Gemma-3-12b-it (Prompt) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 30.43 | percentage | higher_is_better | 66 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
bfcl-gemma-3-1b-it-prompt-function_calling_accuracy-2026-05-17 | Gemma-3-1b-it (Prompt) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 7.17 | percentage | higher_is_better | 109 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.219Z |
bfcl-gemma-3-27b-it-prompt-function_calling_accuracy-2026-05-17 | Gemma-3-27b-it (Prompt) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 29.47 | percentage | higher_is_better | 69 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-gemma-3-4b-it-prompt-function_calling_accuracy-2026-05-17 | Gemma-3-4b-it (Prompt) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 19.62 | percentage | higher_is_better | 101 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.219Z |
bfcl-glm-4-6-fc-thinking-function_calling_accuracy-2026-05-17 | GLM-4.6 (FC thinking) | null | null | Zhipu AI | agente autónomo | fallo técnico | function_calling_accuracy | 72.38 | percentage | higher_is_better | 4 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.213Z |
bfcl-gpt-4-1-2025-04-14-fc-function_calling_accuracy-2026-05-17 | GPT-4.1-2025-04-14 (FC) | null | null | OpenAI | agente autónomo | fallo técnico | function_calling_accuracy | 53.96 | percentage | higher_is_better | 20 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.214Z |
bfcl-gpt-4-1-2025-04-14-prompt-function_calling_accuracy-2026-05-17 | GPT-4.1-2025-04-14 (Prompt) | null | null | OpenAI | agente autónomo | fallo técnico | function_calling_accuracy | 39.38 | percentage | higher_is_better | 45 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
bfcl-gpt-4-1-mini-2025-04-14-fc-function_calling_accuracy-2026-05-17 | GPT-4.1-mini-2025-04-14 (FC) | null | null | OpenAI | agente autónomo | fallo técnico | function_calling_accuracy | 50.45 | percentage | higher_is_better | 27 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.215Z |
bfcl-gpt-4-1-mini-2025-04-14-prompt-function_calling_accuracy-2026-05-17 | GPT-4.1-mini-2025-04-14 (Prompt) | null | null | OpenAI | agente autónomo | fallo técnico | function_calling_accuracy | 29.73 | percentage | higher_is_better | 67 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-gpt-4-1-nano-2025-04-14-fc-function_calling_accuracy-2026-05-17 | GPT-4.1-nano-2025-04-14 (FC) | null | null | OpenAI | agente autónomo | fallo técnico | function_calling_accuracy | 33.05 | percentage | higher_is_better | 58 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
bfcl-gpt-4-1-nano-2025-04-14-prompt-function_calling_accuracy-2026-05-17 | GPT-4.1-nano-2025-04-14 (Prompt) | null | null | OpenAI | agente autónomo | fallo técnico | function_calling_accuracy | 24.88 | percentage | higher_is_better | 90 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-gpt-5-2-2025-12-11-fc-function_calling_accuracy-2026-05-17 | GPT-5.2-2025-12-11 (FC) | null | null | OpenAI | agente autónomo | fallo técnico | function_calling_accuracy | 55.87 | percentage | higher_is_better | 16 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.214Z |
bfcl-gpt-5-2-2025-12-11-prompt-function_calling_accuracy-2026-05-17 | GPT-5.2-2025-12-11 (Prompt) | null | null | OpenAI | agente autónomo | fallo técnico | function_calling_accuracy | 45.27 | percentage | higher_is_better | 38 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.216Z |
bfcl-gpt-5-mini-2025-08-07-fc-function_calling_accuracy-2026-05-17 | GPT-5-mini-2025-08-07 (FC) | null | null | OpenAI | agente autónomo | fallo técnico | function_calling_accuracy | 55.46 | percentage | higher_is_better | 17 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.214Z |
bfcl-gpt-5-mini-2025-08-07-prompt-function_calling_accuracy-2026-05-17 | GPT-5-mini-2025-08-07 (Prompt) | null | null | OpenAI | agente autónomo | fallo técnico | function_calling_accuracy | 27.83 | percentage | higher_is_better | 77 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-gpt-5-nano-2025-08-07-fc-function_calling_accuracy-2026-05-17 | GPT-5-nano-2025-08-07 (FC) | null | null | OpenAI | agente autónomo | fallo técnico | function_calling_accuracy | 51.45 | percentage | higher_is_better | 24 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.214Z |
bfcl-gpt-5-nano-2025-08-07-prompt-function_calling_accuracy-2026-05-17 | GPT-5-nano-2025-08-07 (Prompt) | null | null | OpenAI | agente autónomo | fallo técnico | function_calling_accuracy | 27.55 | percentage | higher_is_better | 79 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-granite-20b-functioncalling-fc-function_calling_accuracy-2026-05-17 | Granite-20b-FunctionCalling (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 23.23 | percentage | higher_is_better | 93 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-granite-3-1-8b-instruct-fc-function_calling_accuracy-2026-05-17 | Granite-3.1-8B-Instruct (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 27.1 | percentage | higher_is_better | 81 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-granite-3-2-8b-instruct-fc-function_calling_accuracy-2026-05-17 | Granite-3.2-8B-Instruct (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 26.87 | percentage | higher_is_better | 83 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-granite-4-0-350m-fc-function_calling_accuracy-2026-05-17 | Granite-4.0-350m (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 18.98 | percentage | higher_is_better | 103 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.219Z |
bfcl-grok-4-0709-fc-function_calling_accuracy-2026-05-17 | Grok-4-0709 (FC) | null | null | xAI | agente autónomo | fallo técnico | function_calling_accuracy | 61.38 | percentage | higher_is_better | 10 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.213Z |
bfcl-grok-4-0709-prompt-function_calling_accuracy-2026-05-17 | Grok-4-0709 (Prompt) | null | null | xAI | agente autónomo | fallo técnico | function_calling_accuracy | 62.97 | percentage | higher_is_better | 9 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.213Z |
bfcl-grok-4-1-fast-non-reasoning-fc-function_calling_accuracy-2026-05-17 | Grok-4-1-fast-non-reasoning (FC) | null | null | xAI | agente autónomo | fallo técnico | function_calling_accuracy | 58.29 | percentage | higher_is_better | 12 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.213Z |
bfcl-grok-4-1-fast-reasoning-fc-function_calling_accuracy-2026-05-17 | Grok-4-1-fast-reasoning (FC) | null | null | xAI | agente autónomo | fallo técnico | function_calling_accuracy | 69.57 | percentage | higher_is_better | 5 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.213Z |
bfcl-hammer2-1-0-5b-fc-function_calling_accuracy-2026-05-17 | Hammer2.1-0.5b (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 21.22 | percentage | higher_is_better | 100 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.219Z |
bfcl-hammer2-1-1-5b-fc-function_calling_accuracy-2026-05-17 | Hammer2.1-1.5b (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 27.88 | percentage | higher_is_better | 75 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-hammer2-1-3b-fc-function_calling_accuracy-2026-05-17 | Hammer2.1-3b (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 29.71 | percentage | higher_is_better | 68 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-hammer2-1-7b-fc-function_calling_accuracy-2026-05-17 | Hammer2.1-7b (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 31.67 | percentage | higher_is_better | 64 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
bfcl-llama-3-1-8b-instruct-prompt-function_calling_accuracy-2026-05-17 | Llama-3.1-8B-Instruct (Prompt) | null | null | Meta | agente autónomo | fallo técnico | function_calling_accuracy | 25.83 | percentage | higher_is_better | 85 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-llama-3-1-nemotron-ultra-253b-v1-fc-function_calling_accuracy-2026-05-17 | Llama-3.1-Nemotron-Ultra-253B-v1 (FC) | null | null | Meta | agente autónomo | fallo técnico | function_calling_accuracy | 10 | percentage | higher_is_better | 108 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.219Z |
bfcl-llama-3-2-1b-instruct-fc-function_calling_accuracy-2026-05-17 | Llama-3.2-1B-Instruct (FC) | null | null | Meta | agente autónomo | fallo técnico | function_calling_accuracy | 10.82 | percentage | higher_is_better | 107 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.219Z |
bfcl-llama-3-2-3b-instruct-fc-function_calling_accuracy-2026-05-17 | Llama-3.2-3B-Instruct (FC) | null | null | Meta | agente autónomo | fallo técnico | function_calling_accuracy | 21.95 | percentage | higher_is_better | 98 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-llama-3-3-70b-instruct-fc-function_calling_accuracy-2026-05-17 | Llama-3.3-70B-Instruct (FC) | null | null | Meta | agente autónomo | fallo técnico | function_calling_accuracy | 31.9 | percentage | higher_is_better | 62 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
bfcl-llama-4-maverick-17b-128e-instruct-fp8-fc-function_calling_accuracy-2026-05-17 | Llama-4-Maverick-17B-128E-Instruct-FP8 (FC) | null | null | Meta | agente autónomo | fallo técnico | function_calling_accuracy | 37.29 | percentage | higher_is_better | 50 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
bfcl-llama-4-scout-17b-16e-instruct-fc-function_calling_accuracy-2026-05-17 | Llama-4-Scout-17B-16E-Instruct (FC) | null | null | Meta | agente autónomo | fallo técnico | function_calling_accuracy | 28.13 | percentage | higher_is_better | 72 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-minicpm3-4b-fc-fc-function_calling_accuracy-2026-05-17 | MiniCPM3-4B-FC (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 25.55 | percentage | higher_is_better | 86 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-minicpm3-4b-prompt-function_calling_accuracy-2026-05-17 | MiniCPM3-4B (Prompt) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 22.08 | percentage | higher_is_better | 97 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-ministral-8b-instruct-2410-fc-function_calling_accuracy-2026-05-17 | Ministral-8B-Instruct-2410 (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 11.1 | percentage | higher_is_better | 105 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.219Z |
bfcl-mistral-large-2411-fc-function_calling_accuracy-2026-05-17 | mistral-large-2411 (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 38.37 | percentage | higher_is_better | 46 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
bfcl-mistral-large-2411-prompt-function_calling_accuracy-2026-05-17 | mistral-large-2411 (Prompt) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 31.84 | percentage | higher_is_better | 63 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
bfcl-mistral-medium-2505-fc-function_calling_accuracy-2026-05-17 | Mistral-Medium-2505 (FC) | null | null | Mistral | agente autónomo | fallo técnico | function_calling_accuracy | 37.56 | percentage | higher_is_better | 49 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
bfcl-mistral-medium-2505-function_calling_accuracy-2026-05-17 | Mistral-Medium-2505 | null | null | Mistral | agente autónomo | fallo técnico | function_calling_accuracy | 37.69 | percentage | higher_is_better | 48 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
bfcl-mistral-small-2506-fc-function_calling_accuracy-2026-05-17 | Mistral-small-2506 (FC) | null | null | Mistral | agente autónomo | fallo técnico | function_calling_accuracy | 37.15 | percentage | higher_is_better | 51 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
bfcl-mistral-small-2506-prompt-function_calling_accuracy-2026-05-17 | Mistral-Small-2506 (Prompt) | null | null | Mistral | agente autónomo | fallo técnico | function_calling_accuracy | 32.38 | percentage | higher_is_better | 59 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
bfcl-moonshotai-kimi-k2-instruct-fc-function_calling_accuracy-2026-05-17 | Moonshotai-Kimi-K2-Instruct (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 59.06 | percentage | higher_is_better | 11 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.213Z |
bfcl-nanbeige3-5-pro-thinking-fc-function_calling_accuracy-2026-05-17 | Nanbeige3.5-Pro-Thinking (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 47.68 | percentage | higher_is_better | 32 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.216Z |
bfcl-nanbeige4-3b-thinking-2511-fc-function_calling_accuracy-2026-05-17 | Nanbeige4-3B-Thinking-2511 (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 51.4 | percentage | higher_is_better | 25 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.215Z |
bfcl-o3-2025-04-16-fc-function_calling_accuracy-2026-05-17 | o3-2025-04-16 (FC) | null | null | OpenAI | agente autónomo | fallo técnico | function_calling_accuracy | 48.56 | percentage | higher_is_better | 30 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.216Z |
bfcl-o3-2025-04-16-prompt-function_calling_accuracy-2026-05-17 | o3-2025-04-16 (Prompt) | null | null | OpenAI | agente autónomo | fallo técnico | function_calling_accuracy | 63.05 | percentage | higher_is_better | 8 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.213Z |
bfcl-o4-mini-2025-04-16-fc-function_calling_accuracy-2026-05-17 | o4-mini-2025-04-16 (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 53.24 | percentage | higher_is_better | 21 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.214Z |
bfcl-o4-mini-2025-04-16-prompt-function_calling_accuracy-2026-05-17 | o4-mini-2025-04-16 (Prompt) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 50.26 | percentage | higher_is_better | 28 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.216Z |
bfcl-open-mistral-nemo-2407-fc-function_calling_accuracy-2026-05-17 | Open-Mistral-Nemo-2407 (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 27.63 | percentage | higher_is_better | 78 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-open-mistral-nemo-2407-prompt-function_calling_accuracy-2026-05-17 | Open-Mistral-Nemo-2407 (Prompt) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 19.31 | percentage | higher_is_better | 102 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.219Z |
bfcl-palmyra-x-004-fc-function_calling_accuracy-2026-05-17 | palmyra-x-004 (FC) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 27.87 | percentage | higher_is_better | 76 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-phi-4-prompt-function_calling_accuracy-2026-05-17 | Phi-4 (Prompt) | null | null | Otro | agente autónomo | fallo técnico | function_calling_accuracy | 28.79 | percentage | higher_is_better | 70 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-qwen3-0-6b-fc-function_calling_accuracy-2026-05-17 | Qwen3-0.6B (FC) | null | null | Alibaba | agente autónomo | fallo técnico | function_calling_accuracy | 23.93 | percentage | higher_is_better | 92 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-qwen3-0-6b-prompt-function_calling_accuracy-2026-05-17 | Qwen3-0.6B (Prompt) | null | null | Alibaba | agente autónomo | fallo técnico | function_calling_accuracy | 22.38 | percentage | higher_is_better | 94 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-qwen3-1-7b-fc-function_calling_accuracy-2026-05-17 | Qwen3-1.7B (FC) | null | null | Alibaba | agente autónomo | fallo técnico | function_calling_accuracy | 28.41 | percentage | higher_is_better | 71 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.218Z |
bfcl-qwen3-14b-fc-function_calling_accuracy-2026-05-17 | Qwen3-14B (FC) | null | null | Alibaba | agente autónomo | fallo técnico | function_calling_accuracy | 41.03 | percentage | higher_is_better | 43 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
bfcl-qwen3-14b-prompt-function_calling_accuracy-2026-05-17 | Qwen3-14B (Prompt) | null | null | Alibaba | agente autónomo | fallo técnico | function_calling_accuracy | 37.77 | percentage | higher_is_better | 47 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
bfcl-qwen3-235b-a22b-instruct-2507-fc-function_calling_accuracy-2026-05-17 | Qwen3-235B-A22B-Instruct-2507 (FC) | null | null | Alibaba | agente autónomo | fallo técnico | function_calling_accuracy | 47.99 | percentage | higher_is_better | 31 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.216Z |
bfcl-qwen3-235b-a22b-instruct-2507-prompt-function_calling_accuracy-2026-05-17 | Qwen3-235B-A22B-Instruct-2507 (Prompt) | null | null | Alibaba | agente autónomo | fallo técnico | function_calling_accuracy | 52.15 | percentage | higher_is_better | 23 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.214Z |
bfcl-qwen3-30b-a3b-instruct-2507-fc-function_calling_accuracy-2026-05-17 | Qwen3-30B-A3B-Instruct-2507 (FC) | null | null | Alibaba | agente autónomo | fallo técnico | function_calling_accuracy | 41.39 | percentage | higher_is_better | 41 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.216Z |
bfcl-qwen3-30b-a3b-instruct-2507-prompt-function_calling_accuracy-2026-05-17 | Qwen3-30B-A3B-Instruct-2507 (Prompt) | null | null | Alibaba | agente autónomo | fallo técnico | function_calling_accuracy | 36.7 | percentage | higher_is_better | 53 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
bfcl-qwen3-32b-fc-function_calling_accuracy-2026-05-17 | Qwen3-32B (FC) | null | null | Alibaba | agente autónomo | fallo técnico | function_calling_accuracy | 48.71 | percentage | higher_is_better | 29 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.216Z |
bfcl-qwen3-32b-prompt-function_calling_accuracy-2026-05-17 | Qwen3-32B (Prompt) | null | null | Alibaba | agente autónomo | fallo técnico | function_calling_accuracy | 46.78 | percentage | higher_is_better | 33 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.216Z |
bfcl-qwen3-4b-instruct-2507-fc-function_calling_accuracy-2026-05-17 | Qwen3-4B-Instruct-2507 (FC) | null | null | Alibaba | agente autónomo | fallo técnico | function_calling_accuracy | 35.68 | percentage | higher_is_better | 54 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
bfcl-qwen3-4b-instruct-2507-prompt-function_calling_accuracy-2026-05-17 | Qwen3-4B-Instruct-2507 (Prompt) | null | null | Alibaba | agente autónomo | fallo técnico | function_calling_accuracy | 35.52 | percentage | higher_is_better | 55 | bfcl | https://gorilla.cs.berkeley.edu/leaderboard.html | Apache-2.0 | https://gorilla.cs.berkeley.edu/leaderboard.html | 2026-05-17T23:40:13.328Z | 2026-05-17T09:54:00.217Z |
- Cifras del snapshot actual
- Para que sirve este dataset
- Snapshots disponibles
- Que contiene cada snapshot
- Estructura del dataset (18 columnas estables del CSV)
- Taxonomia cerrada
- Ejemplo de fila real del snapshot actual
- Fuentes oficiales
- Como usar
- Considerations for Using the Data
- Equivalencia con la API publica
- Como citar
- Sitio web y otras distribuciones
- Dataset Curators
- Contributions
- Mantenimiento y contacto
- Licencia
Indice de Fallos IA en espanol
Snapshots mensuales del Indice de Fallos IA producido por el observatorio La AutopsIA (ApisDom Intelligence Group). Mide la fiabilidad de modelos LLM con benchmarks oficiales independientes, en formato citable y trazable.
Cifras del snapshot actual
- 678 mediciones en este snapshot.
- Mes archivado: 2026-05.
- Recomputado: 2026-05-17T23:42:53.152Z.
- Frecuencia: sincronizacion mensual.
Para que sirve este dataset
Datos crudos, comparables y auditables de la fiabilidad real de modelos LLM. Casos de uso:
- Equipos de producto eligiendo que LLM integrar: comparar tasa de alucinacion, precision en function calling y rating Elo entre candidatos con datos verificables, no marketing del proveedor.
- Investigacion academica sobre evolucion temporal de la fiabilidad LLM: snapshots mensuales encadenados permiten estudios longitudinales con cita por DOI.
- Periodismo tecnico y divulgacion: cifras citables con DOI permanente, no leaderboards efimeros.
- Auditoria y compliance: trazabilidad fila a fila (
sourceUrl,sourceLicense,methodology,measuredAt,fetchedAt) para reconstruir cualquier medicion hasta su fuente. - Reguladores evaluando criterios objetivos sobre fiabilidad de sistemas IA generativa.
No es un ranking de calidad general de un modelo. Cada metrica mide un aspecto concreto.
Snapshots disponibles
| Mes | JSON | CSV | DOI Zenodo |
|---|---|---|---|
| 2026-05 | json | csv | 10.5281/zenodo.20218106 |
La columna DOI Zenodo apunta al DOI vigente del snapshot agregador. Cada snapshot mensual emite su propio DOI permanente.
Que contiene cada snapshot
Cada mes se publican dos ficheros con el mismo contenido en dos formatos:
indice-fallos-ia-YYYY-MM.json: envoltura{ meta, data }, dondedataes elBenchmarkLiveStatecompleto agrupado por metrica. Las filas dentro dedata.byMetric.*.rowssiguen el tipo publicoPublicBenchmarkResult(18 campos, ver siguiente seccion). Claves cuyos valores no aporta la fuente se OMITEN del JSON.indice-fallos-ia-YYYY-MM.csv: filas planas, 18 columnas estables (RFC 4180 con BOM UTF-8, encapsulado total entre comillas dobles), una fila por (modelo x metrica). Celdas sin valor quedan vacias (par de comillas vacias).
El README.md se sobrescribe cada mes con las cifras del snapshot vigente.
Estructura del dataset (18 columnas estables del CSV)
| Columna | Tipo | Descripcion |
|---|---|---|
| id | string | Identificador unico del registro. |
| modelVariant | string | Identificador del modelo (texto libre del leaderboard). |
| modelVersion | string opcional | Version del modelo. Si la fuente no la aporta, celda vacia y clave JSON omitida. |
| aiSystem | string opcional | Sistema IA canonico via alias. |
| aiCompany | string opcional | Empresa proveedora (taxonomia cerrada). |
| aiModality | string opcional | Modalidad del sistema (taxonomia cerrada). |
| failureType | string opcional | Tipo de fallo (taxonomia cerrada). |
| metricName | string | Nombre tecnico de la metrica (taxonomia cerrada). |
| metricValue | number | Valor numerico medido. |
| metricUnit | string | Unidad (percentage, score_0_1, elo, count). |
| metricDirection | string | higher_is_better o lower_is_better. |
| rank | number opcional | Rango del modelo en su metrica (1 es mejor). |
| source | string | Fuente original (vectara, bfcl, lmsys, ...). |
| sourceUrl | string | URL canonica de la fuente. |
| sourceLicense | string | Licencia de la fuente (taxonomia cerrada). |
| methodology | string | URL de la metodologia documentada. |
| measuredAt | string ISO | Cuando se midio en origen. |
| fetchedAt | string ISO | Cuando lo recogio La AutopsIA. |
Regla del dataset (cero invento): si la fuente original no aporta un campo, la celda CSV queda vacia (par de comillas vacias RFC 4180) y la clave correspondiente NO aparece en el JSON. No se rellena con null, ni con placeholders (N/A, desconocido, pendiente). El consumidor distingue tres estados sin ambiguedad: clave ausente (sin dato), clave con string vacio (vacio explicito), clave con valor (dato presente).
Taxonomia cerrada
Los campos categoricos del dataset solo toman valores de listas cerradas declaradas en codigo (src/config/benchmark-taxonomy.ts). No hay valores libres en estos campos.
- Fuentes (source):
vectara,bfcl,lmsys,simpleqa,open-llm-leaderboard. - Metricas (metricName):
hallucination_rate,factual_consistency,function_calling_accuracy,function_calling_simple,function_calling_multiple,function_calling_parallel,elo_rating,mmlu_score,gsm8k_score,truthfulqa_score. - Unidades (metricUnit):
percentage,score_0_1,elo,count. - Direcciones (metricDirection):
higher_is_better,lower_is_better. - Licencias permitidas (sourceLicense):
Apache-2.0,CC-BY-4.0,CC-BY-SA-4.0,MIT,Public-Domain. - Estado de aprobacion (approvalStatus):
pending,approved,rejected.
Reglas de evolucion: se pueden anadir entradas en el futuro sin romper datos guardados; nunca se renombran ni eliminan mientras existan documentos que las usen.
Ejemplo de fila real del snapshot actual
CSV (las dos primeras lineas: cabecera y la primera fila de datos, tal como aparecen en el fichero):
"id","modelVariant","modelVersion","aiSystem","aiCompany","aiModality","failureType","metricName","metricValue","metricUnit","metricDirection","rank","source","sourceUrl","sourceLicense","methodology","measuredAt","fetchedAt"
"bfcl-amazon-nova-2-lite-v1-0-fc-function_calling_accuracy-2026-05-17","Amazon-Nova-2-Lite-v1:0 (FC)","","","Otro","agente autónomo","fallo técnico","function_calling_accuracy","27.1","percentage","higher_is_better","80","bfcl","https://gorilla.cs.berkeley.edu/leaderboard.html","Apache-2.0","https://gorilla.cs.berkeley.edu/leaderboard.html","2026-05-17T23:40:13.328Z","2026-05-17T09:54:00.218Z"
JSON (la primera fila plana, tal como aparece dentro de data.byMetric.{metric}.rows[0] del fichero JSON):
{
"id": "bfcl-amazon-nova-2-lite-v1-0-fc-function_calling_accuracy-2026-05-17",
"modelVariant": "Amazon-Nova-2-Lite-v1:0 (FC)",
"metricName": "function_calling_accuracy",
"metricValue": 27.1,
"metricUnit": "percentage",
"metricDirection": "higher_is_better",
"source": "bfcl",
"sourceUrl": "https://gorilla.cs.berkeley.edu/leaderboard.html",
"sourceLicense": "Apache-2.0",
"measuredAt": "2026-05-17T23:40:13.328Z",
"fetchedAt": "2026-05-17T09:54:00.218Z",
"methodology": "https://gorilla.cs.berkeley.edu/leaderboard.html",
"aiCompany": "Otro",
"aiModality": "agente autónomo",
"failureType": "fallo técnico",
"rank": 80
}
Ambos ejemplos son la primera fila REAL del snapshot vigente. Si esta es la primera vez que se publica un snapshot, los valores cambian al volver a sincronizar.
Fuentes oficiales
| Fuente | URL | Metricas que aporta | Licencia |
|---|---|---|---|
| Vectara Hallucination Leaderboard | https://github.com/vectara/hallucination-leaderboard | hallucination_rate, factual_consistency |
Apache-2.0 |
| BFCL (Berkeley Function Calling Leaderboard) | https://gorilla.cs.berkeley.edu/leaderboard.html | function_calling_accuracy y variantes |
Apache-2.0 |
| LMSYS Chatbot Arena (mirror diario) | https://huggingface.co/spaces/lmsys/chatbot-arena-leaderboard | elo_rating |
MIT |
Si en el futuro se reactivan fuentes hoy desactivadas (SimpleQA, Open LLM Leaderboard), apareceran automaticamente con sus metricas correspondientes sin tocar el esquema.
Como usar
CSV con la libreria datasets:
from datasets import load_dataset
ds = load_dataset("apisdom/indice-fallos-ia-espanol", data_files="indice-fallos-ia-2026-05.csv")
JSON crudo via HTTP:
import json, urllib.request
url = "https://huggingface.co/datasets/apisdom/indice-fallos-ia-espanol/resolve/main/indice-fallos-ia-2026-05.json"
with urllib.request.urlopen(url) as r:
snap = json.load(r)
print(snap["meta"], len(snap["data"]["byMetric"]))
Considerations for Using the Data
Discussion of Biases
- Sesgo de seleccion: solo se publican benchmarks de fuentes con leaderboard publico documentado. Modelos evaluados solo en benchmarks privados quedan fuera.
- Sesgo de mercado y de idioma: las fuentes activas (Vectara, BFCL, LMSYS) tienden a cubrir mejor modelos comerciales en ingles que modelos en otros idiomas o procedentes de investigacion academica menos visible.
- Sesgo metodologico: cada fuente mide aspectos distintos (alucinacion, function calling, preferencia humana). El ranking agregado no refleja calidad general de un modelo, solo su rendimiento en el aspecto medido. Vectara mide alucinacion con prompts fijos; BFCL evalua function calling sintetico; LMSYS depende de votaciones humanas. Estos sesgos se heredan en el dataset y no se intentan corregir.
Other Known Limitations
- Cobertura no exhaustiva: el snapshot refleja unicamente los modelos LLM que aparecen en las fuentes activas al cierre del mes. Modelos comerciales privados sin evaluacion publica no aparecen, no por ser peores, sino porque no hay dato.
- Ventana temporal de medicion:
measuredAtindica cuando se midio en origen,fetchedAtcuando lo recogio este sistema. Hay desfases naturales de horas o dias. - Datos faltantes: si una fuente no aporta un campo, la celda CSV queda vacia y la clave JSON se omite.
- El ranking depende de los modelos disponibles ese mes: comparar entre meses requiere alinear por
modelVariant, no porrank.
Recommended and Non-Recommended Uses
- Recomendado: auditoria comparativa de fiabilidad de LLM, investigacion academica sobre evolucion temporal, periodismo tecnico y divulgacion sobre seguridad de IA.
- NO recomendado: toma de decisiones clinicas, legales o financieras basadas directamente en el ranking. El dataset mide aspectos especificos, no aptitud general para tareas reguladas.
- NO recomendado: training de modelos. Es dataset de evaluacion, no de entrenamiento.
Personal and Sensitive Information
El dataset no contiene datos personales identificables.
Equivalencia con la API publica
Los ficheros de este dataset son identicos a los que sirven los endpoints publicos del observatorio:
- https://laautopsia.com/api/dataset/v1/benchmarks/live.json
- https://laautopsia.com/api/dataset/v1/benchmarks/live.csv
Como citar
@dataset{laautopsia_indice_2026_05,
author = {Salvador Valdivieso, Juan Luis},
title = {Indice de Fallos IA - Snapshot 2026-05},
year = {2026},
publisher = {Zenodo},
organization = {ApisDom Intelligence Group, laautopsia.com},
version = {2026-05},
doi = {10.5281/zenodo.20218106},
url = {https://huggingface.co/datasets/apisdom/indice-fallos-ia-espanol}
}
Cada snapshot mensual emite un DOI nuevo permanente.
Sitio web y otras distribuciones
- https://laautopsia.com: observatorio con dashboard, RSS y API publica.
- Zenodo: snapshots mensuales con DOI permanente citable.
- datos.gob.es: publicacion prevista en plan de visibilidad.
Dataset Curators
Mantenido por ApisDom Intelligence Group (apisdom.com) a traves del observatorio La AutopsIA (https://laautopsia.com). El equipo editorial define la taxonomia cerrada, valida cada medicion publicada y mantiene los scrapers de las fuentes oficiales.
Contributions
Las contribuciones se gestionan a traves del panel de discusiones de este dataset en Hugging Face o por email a dataset@apisdom.com. Las aportaciones aceptadas se publican en el siguiente snapshot mensual con atribucion al autor en el changelog. Tipos de contribucion bienvenidos: deteccion de errores en mediciones, fuentes oficiales nuevas que cumplan los criterios de licencia y metodologia publica, mejoras a la taxonomia.
Mantenimiento y contacto
- Frecuencia: snapshot mensual, dia 3 a las 09:00 UTC.
- Mantenedor: ApisDom Intelligence Group - apisdom.com.
- Observatorio: laautopsia.com.
- Contacto del dataset: dataset@apisdom.com.
- Contacto editorial: noticias-autopsia@apisdom.com.
- Issues tecnicos: panel de discusiones de este dataset en Hugging Face.
Licencia
Este dataset se publica bajo CC BY-SA 4.0. Las licencias originales de cada fuente se preservan en el campo sourceLicense de cada fila: Apache-2.0 para Vectara y BFCL; MIT para LMSYS (el dato proviene del mirror diario MIT de la comunidad, no del leaderboard original que es CC-BY-4.0).
- Downloads last month
- 52