Nejumi 4 - BFCL - Multi-Turn: leaderboard
Berkeley Function Calling Leaderboard as adapted and sampled by Nejumi Leaderboard 4 (Weights & Biases Japan): accuracy on the multi-turn base, missing-function and missing-parameter categories. Accuracy (%), higher is better; the overall accuracy is Nejumi 4 - GLP - Function Calling.
Metric: Accuracy (%). Source: nejumi.ai. Saturation forecast: Around August 2027. 145 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.6 Sol (Max) | 66.67 | #22 (GPT-5.6 Sol) |
| 2 | Claude Opus 4.5 (Thinking) | 66.67 | #87 (Claude Opus 4.5) |
| 3 | Claude Opus 4.7 (xHigh) | 66.67 | #51 (Claude Opus 4.7) |
| 4 | Claude Opus 4.1 (Thinking) | 66.67 | #146 (Claude Opus 4.1) |
| 5 | Claude Opus 5 (Max) | 65.56 | #16 (Claude Opus 5) |
| 6 | Claude Opus 4.5 (Non-reasoning) | 65.56 | #87 (Claude Opus 4.5) |
| 7 | Claude Opus 4.6 (Thinking) | 64.44 | #67 (Claude Opus 4.6) |
| 8 | Claude Opus 4.8 (xHigh) | 64.44 | #39 (Claude Opus 4.8) |
| 9 | GPT-4o Mini (2024-07-18) | 61.67 | #568 |
| 10 | Claude Sonnet 4.6 (Thinking) | 61.11 | #94 (Claude Sonnet 4.6) |
| 11 | MiniMax-M3 (Thinking) | 61.11 | #151 (MiniMax-M3) |
| 12 | Claude Opus 4 (Non-reasoning) | 60 | #170 (Claude Opus 4) |
| 13 | Claude 3.5 Sonnet | 58.89 | #362 |
| 14 | GPT-5.6 Luna (Max) | 58.89 | #79 (GPT-5.6 Luna) |
| 15 | Grok 3 | 56.67 | #313 |
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-multi-turn · How It Works · Data refreshed daily, snapshot 2026-11-05.