Olmo 3 — local AI model by Ai2 (Allen Institute for AI)
Written by Jakub Rusinowski · Last updated
Fully open, not just open-weight: Ai2 publishes the weights, the training data (Dolma 3), the code and the intermediate checkpoints under Apache-2.0. Base, Instruct and Think variants at each size, which makes this the family to reach for when the training process itself has to be auditable.
Variants
The smallest Olmo 3 variant needs about 5 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache.
| Model | VRAM at Q4 | VRAM | Context | Run it |
|---|---|---|---|---|
| Olmo 3 7B → 7B | ~5 GB | 65,536 | ollama run olmo-3 | |
| Olmo 3 32B → 32B | ~20.1 GB | 65,536 | ollama run olmo-3 |
Memory is quantized weights plus overhead at Q4_K_M, from the same engine as the GPU & VRAM checker.
How to run Olmo 3 locally
Install Ollama, then pull the tag.
ollama run olmo-3Pick a size above for its own VRAM figure, speed estimate and install command.
Licence
Commercial use permitted. No usage restrictions beyond attribution.
Applies to: Olmo 3 7B, Olmo 3 32BRecommended GPU
The cheapest catalogued GPU that runs Olmo 3 locally (min 5 GB VRAM) is the Intel Arc B570 (10 GB).
Olmo 3 — frequently asked questions
How much VRAM does Olmo 3 need?
Olmo 3 needs about 5 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Olmo 3 7B (5 GB, Q4_K_M); Olmo 3 32B (20 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run Olmo 3 on an RTX 4090 (24 GB)?
Yes — Olmo 3 runs on an RTX 4090 (24 GB) and other 24 GB cards such as the RTX 3090. Smaller variants also fit comfortably on 8–16 GB GPUs at Q4_K_M.
What quantization should I use for Olmo 3?
Q4_K_M is the best balance of quality and VRAM for Olmo 3 in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
How do I run Olmo 3 with Ollama?
Olmo 3 has no local Ollama tag — the published tag is cloud-hosted, so running it sends your prompts to a hosted GPU rather than your own machine.