Olmo 3 32B Think

About 23 GB at Q4, 39.9 GB at Q8 and 74.6 GB at FP16. All figures are estimates from the 32.2B parameter count, weights plus about 15% and 0.5 GB for a short context.

Yes on a 24 GB GPU, which fits at Q4 (estimated 23 GB). It does not fit at Q4, Q8 or FP16 on 16 GB, 12 GB or 8 GB GPUs (estimate).

Yes, at Q4 (estimated 23 GB). Assuming 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 3 32B Think has open weights, so you can self-host it for full control of your data.

Run Olmo 3 32B Think locally

No. Input modalities are text only, so it does not read images, audio, video or PDFs.

Rank 284 on Arena Text, with a score of 1306.

Olmo 3 32B Think benchmarks

Closest alternatives, matched on model kind and release date (no comparable price): ERNIE 4.5 VL 28B A3B Thinking by Baidu, DeepSeek Reasoner by DeepSeek and Trinity Mini by Arcee AI.

ERNIE 4.5 VL 28B A3B ThinkingOlmo 3 32B Think vs ERNIE 4.5 VL 28B A3B ThinkingDeepSeek ReasonerOlmo 3 32B Think vs DeepSeek ReasonerTrinity MiniOlmo 3 32B Think vs Trinity Mini

Olmo 3 32B Think has open weights, licensed under Apache 2.0.

Yes. The Apache 2.0 license is a permissive license that allows commercial use.

You can download the weights from the Hugging Face repository allenai/Olmo-3-32B-Think.

Run Olmo 3 32B Think locally

Olmo 3 32B Think was made by Ai2 and released on November 20, 2025.

More from Ai2

Yes. Olmo 3 32B Think has open weights.