NVIDIA Nemotron 49B v1.5 Card - BFCL v3: leaderboard
Berkeley Function Calling Leaderboard v3 function-calling accuracy, which the card reports as pass@1 averaged over 2 runs. From NVIDIA's own Llama-3.3-Nemotron-Super-49B-v1.5 model card, which prints only a reasoning-on result and states evaluation with NeMo-Skills at temperature 0.6, top-p 0.95 and a 64k sequence length. NVIDIA's separate NeMo-Skills reproduction tutorial reports its own 16-run numbers on different boards.
Metric: pass@1 (avg. over 2 runs). Source: huggingface.co. 1 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama-3.3-nemotron-super-49B-v1.5 (Reasoning) | 71.75 | #903 (Llama-3.3-nemotron-super-49B-v1.5) |
Interactive version: theaggregate.ai/benchmark?slug=nvidia-nemotron-49b-v1-5-card-bfcl-v3 · How It Works · Data refreshed daily, snapshot 2026-11-12.