Nejumi 4 - BFCL - Live AST: leaderboard
Berkeley Function Calling Leaderboard as adapted and sampled by Nejumi Leaderboard 4 (Weights & Biases Japan): abstract-syntax-tree accuracy on the live (user-contributed) simple and multiple-function categories. Accuracy (%), higher is better; the overall accuracy is Nejumi 4 - GLP - Function Calling.
Metric: Accuracy (%). Source: nejumi.ai. Saturation forecast: Not forecast. 145 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 122B A10B (Thinking) | 81.48 | #184 (Qwen 3.5 122B A10B) |
| 2 | Qwen 3.5 397B A17B (Thinking) | 76.85 | #154 (Qwen 3.5 397B A17B) |
| 3 | Qwen 3.6 Max (Preview) (Thinking) | 76.85 | |
| 4 | Qwen 3.8 27B (xHigh) | 75.93 | #123 (Qwen 3.8 27B) |
| 5 | Qwen 3.6 27B (Thinking) | 75.93 | #182 (Qwen 3.6 27B) |
| 6 | Qwen3-Swallow-32B-RL-v0.2 (Thinking) | 75.93 | #374 (Qwen3-Swallow-32B-RL-v0.2) |
| 7 | GPT-OSS-Swallow-120B-RL-v0.1 (Thinking) | 75 | #309 (GPT-OSS-Swallow-120B-RL-v0.1) |
| 8 | Gemini 3.1 Pro (Preview) | 74.07 | #64 |
| 9 | Gemini 2.5 Flash | 74.07 | #255 |
| 10 | Kimi K3 (Max) | 74.07 | #35 (Kimi K3) |
| 11 | Grok 4.1 Fast | 74.07 | #227 |
| 12 | DeepSeek V4 Pro (Max) | 74.07 | #112 (DeepSeek V4 Pro) |
| 13 | Qwen 3.5 4B (Thinking) | 74.07 | #514 (Qwen 3.5 4B) |
| 14 | Qwen 3.5 9B (Thinking) | 74.07 | #391 (Qwen 3.5 9B) |
| 15 | Qwen 3.5 27B (Thinking) | 74.07 | #192 (Qwen 3.5 27B) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=nejumi-4-bfcl-live-ast · How It Works · Data refreshed daily, snapshot 2026-11-05.