Molmo 2 — local AI model by Allen Institute for AI (Ai2)
Written by Jakub Rusinowski · Last updated
Ai2's open vision-language models for images, multiple images and video, including pointing at and tracking objects. Weights, training data and code are published. Molmo 2 8B is built on Qwen3-8B with a SigLIP 2 vision encoder.
Variants
The smallest Molmo 2 variant needs about 6 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache.
| Model | VRAM at Q4 | VRAM | Context | Run it |
|---|---|---|---|---|
| Molmo 2 8B → 8B | ~6 GB | 36,864 | ollama run molmo |
Memory is quantized weights plus overhead at Q4_K_M, from the same engine as the GPU & VRAM checker.
How to run Molmo 2 locally
Install Ollama, then pull the tag.
ollama run molmoPick a size above for its own VRAM figure, speed estimate and install command.
Licence
Commercial use permitted. No usage restrictions beyond attribution.
Applies to: Molmo 2 8BRecommended GPU
The cheapest catalogued GPU that runs Molmo 2 locally (min 6 GB VRAM) is the Intel Arc B570 (10 GB).
Molmo 2 — frequently asked questions
How much VRAM does Molmo 2 need?
Molmo 2 needs about 6 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Molmo 2 8B (6 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run Molmo 2 on an RTX 4090 (24 GB)?
Yes — Molmo 2 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 Molmo 2?
Q4_K_M is the best balance of quality and VRAM for Molmo 2 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 Molmo 2 with Ollama?
Molmo 2 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.