OLMo 2 32B

At Q4 OLMo 2 32B needs an estimated 23 GB of VRAM, at Q8 an estimated 39.9 GB and at FP16 an estimated 74.6 GB. These are estimates from the 32.2B parameter count, covering weights plus about 15% and 0.5 GB for a short context.

Yes, on a 24 GB GPU OLMo 2 32B fits at Q4 (estimated 23 GB). It does not fit on 16 GB, 12 GB or 8 GB GPUs at Q4, Q8 or FP16, based on estimates.

Yes. Based on the estimated 23 GB needed at Q4 and the assumption that macOS lets the GPU use about 75% of unified memory, the smallest common Mac memory size that fits Q4 is 32 GB (estimate).

Yes, OLMo 2 32B has open weights, so you can self-host it for full control of your data.

Run OLMo 2 32B locally

No, OLMo 2 32B supports only text input.

Matched on model kind and release date (no comparable price), the closest alternatives are Command A by Cohere, Llama 3.3 Nemotron Super 49B v1 by NVIDIA and QwQ 32B by Alibaba.

Command AOLMo 2 32B vs Command ALlama 3.3 Nemotron Super 49B v1OLMo 2 32B vs Llama 3.3 Nemotron Super 49B v1QwQ 32BOLMo 2 32B vs QwQ 32B

OLMo 2 32B is open weights, released under the Apache 2.0 license.

Yes, OLMo 2 32B uses the Apache 2.0 license, a permissive license that allows commercial use.

Download the OLMo 2 32B weights from the Hugging Face repository allenai/OLMo-2-0325-32B-Instruct.

Run OLMo 2 32B locally

OLMo 2 32B was made by Ai2 and released on March 13, 2025.

More from Ai2

Yes, OLMo 2 32B is open weights.