Skip to main contentThe Local Bench

Math and reasoning · NVIDIA

OpenReasoning Nemotron 32B

Hugging Face: nvidia/OpenReasoning-Nemotron-32B · checked 2026-10-13

Qwen2.5-32B tuned on R1-0528 maths, code and science traces. Needs 24 GB at Q4.

Datasheet

Maker
NVIDIA
Parameters
32.76 BVerifiedsafetensors.total from the HF API Source
Licence
CC-BY-4.0Verified
Access
Public download
Task (HF pipeline)
text-generation
First published on HF
2025-07-15
Last updated on HF
2025-09-16
Weights in memory
66 GB at 16-bit · 20 GB at ~4.85 bitsEstimateParameters × bits ÷ 8. Excludes context cache and runtime overhead; not every model has a 4-bit build.

Benchmark

89.2%
AIME 2024, pass@1 · #1 of 7 scored in this category

AIME24 column, 32B row, model card. Source

Other published results

  • AIME 2025, pass@1: 84%Sourced Source

See the full math and reasoning ranking

Used in

The description summarises the model card at https://huggingface.co/nvidia/OpenReasoning-Nemotron-32B. Parameters and licence are read from the Hugging Face API; the weights-in-memory line is our arithmetic, not a measurement.