Olmo 3 models

8 Olmo 3 models from Ai2 on Hugging Face, from 7.3B to 32.2B parameters, published by Ai2 in BF16, with a 64K-token context. The smallest official model, Olmo-3-1025-7B, needs about 16.3 GB of VRAM at its published precision; the cheapest live fit is RTX A5000 at $0.176/hr.

Part of the OLMo series · Previous generation: OLMo 2

Pick a size

One row per official Olmo 3 size: the VRAM it needs at each precision, the cheapest GPU that holds it today, and the KV cache for a 32K-token context.

ModelParametersNative VRAMFP8 VRAMINT4 VRAMLive GPU fit (native)Est. $/hrKV cache at 32K
Olmo-3-7B-Instruct7.3B16.3 GB8.2 GB4.1 GBRTX A5000$0.176/hr5.5 GB
Olmo-3-1125-32B32.2B72.0 GB36.0 GB18.0 GBA100$1.31/hr2.8 GB

VRAM is weights times a flat 1.2 overhead; the KV cache is a separate per-model figure at 16-bit, shown where the architecture is published. See the methodology.

Official models (5)

ModelParametersPublished asVRAM neededLive GPU fitEst. $/hr
Olmo-3-1025-7B7.3BBF1616.3 GBRTX A5000$0.176/hr
Olmo-3-7B-Instruct7.3BBF1616.3 GBRTX A5000$0.176/hr
Olmo-3-7B-Think7.3BBF1616.3 GBRTX A5000$0.176/hr
Olmo-3-1125-32B32.2BBF1672.0 GBA100$1.31/hr
Olmo-3.1-32B-Instruct32.2BBF1672.0 GBA100$1.31/hr

Fine-tunes and community models (3)

ModelParametersPublished asVRAM neededLive GPU fitEst. $/hr
Olmo-3-7B-Instruct-DPO7.3BBF1616.3 GBRTX A5000$0.176/hr
Olmo-3-7B-Instruct-SFT7.3BBF1616.3 GBRTX A5000$0.176/hr
Olmo-3-7B-Think-SFT7.3BBF1616.3 GBRTX A5000$0.176/hr

VRAM is for the precision the model is published in: the weights times a flat 1.2 overhead, with the KV cache not included. See the methodology. The fit shown is the lowest-priced single GPU type that holds the model at that precision, or the lowest-priced multi-GPU set (up to 8) when none does.

Run and fine-tune Olmo 3

  • How to run Olmo 3: engine support, chat template and context flags, with a serving command sized per model.
  • How to fine-tune Olmo 3: LoRA and QLoRA memory per model, and the license limits on derivatives.

More from Ai2

Start ComfyUI on a rented GPU.

© 2026 Aquanode. All rights reserved.

All trademarks, logos and brand names are the property of their respective owners.