Pokee-Isaac 28B Report - BFCL v4 - Multiple: leaderboard
Berkeley Function-Calling Leaderboard v4 Multiple category: choosing the right function from several candidates and filling its arguments (200 samples). Score (%). From Pokee AI's Pokee-Isaac 28B technical report (September 10, 2026 revision): Pokee ran every listed model itself through hosted APIs under one harness, with each peer's default temperature (unless a benchmark section fixes one) and a high thinking level where available; exhausted retries, context overflows and verifier failures count as zero. Higher is better.
Metric: Score (%). Source: pokee.ai. Saturation forecast: Estimated already saturated. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 122B A10B | 96.5 | #329 |
| 2 | Nemotron 3 Super | 96 | #766 |
| 3 | Claude Haiku 4.5 | 95 | #568 |
| 4 | Gemini 3.5 Flash Lite | 93.5 | #300 |
| 5 | GPT-5.6 Luna | 89.5 | #117 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=pokee-isaac-28b-report-bfcl-v4-multiple · How It Works · Data refreshed daily, snapshot 2026-11-12.