# Molmo 2 — weights, license and what is actually open

A family of open vision-language models for images, multiple images, and video, with explicit spatial outputs such as pointing and object tracking.

Canonical: https://brightaifuture.com/open-models/molmo
Format: model-family
Source publication: Not established
Bright publication: 2026-09-19
Substantive update: None recorded
Evidence and review: AI-assisted primary-source check on 2026-09-19; no independent model reproduction or license certification. This record does not attach an independent reproduction of Molmo 2 results; it highlights that the released artifacts support such work.

## Documented release

Molmo2-4B, Molmo2-8B, and Molmo2-O-7B

Organization: Allen Institute for AI (Ai2)

Molmo 2 release (month documented; day not established): 2025-12 (month precision). Source: https://allenai.org/media-center

Ai2’s primary media record establishes December 2025, not a release day; this field remains null rather than inventing one.

Parameters: 4B, 7B, and 8B variants

Architecture: Vision-language models with variants specialized for visual understanding and spatial grounding.

Modalities: text, image, video, pointing coordinates, tracking coordinates

Context: Varies by checkpoint and training stage; use the versioned model card.

## License and commercial use

Apache-2.0 repository and model terms; some source datasets require separate agreements

The released code and model artifacts use Apache-2.0, but dataset terms must be reviewed separately.

## What is open

Weights: available. Multiple official checkpoints are released after pre-training, supervised fine-tuning, and long-context fine-tuning. Source: https://github.com/allenai/molmo2

Architecture: available. The repository and technical report document model design and variant differences. Source: https://github.com/allenai/molmo2

Inference code: available. Inference and demo code is included in the official repository. Source: https://github.com/allenai/molmo2

Training code: available. Training scripts for the documented stages are public. Source: https://github.com/allenai/molmo2

Training recipe: available. The repository documents staged training and links the technical report. Source: https://github.com/allenai/molmo2

Data information: available. Training datasets and stage-specific inputs are identified. Source: https://github.com/allenai/molmo2

Training data: partial. Most dataset download scripts are supplied, while some datasets require separate agreements. Source: https://github.com/allenai/molmo2

Evaluation: available. Evaluation code and benchmark details are released with the repository. Source: https://github.com/allenai/molmo2

Commercial use: partial. Apache-2.0 covers released artifacts, while source datasets retain separate terms. Source: https://github.com/allenai/molmo2

## Uses and strengths described in sources

Video understanding

Multi-image reasoning

Pointing

Object tracking

Open training scripts

## Hardware and quantization

Requirements vary from the 4B to 8B variants and with video length; no single minimum is asserted.

No canonical quantized checkpoint is asserted in this record

## Independent evidence

This record does not attach an independent reproduction of Molmo 2 results; it highlights that the released artifacts support such work.

## Limitations

Some training datasets need separate agreements

Spatial outputs still require task-specific evaluation before deployment

## Provenance and history

{}

## Original sources

- [Ai2 media center — Molmo 2 release month](https://allenai.org/media-center)
- [Molmo 2 repository](https://github.com/allenai/molmo2)
- [Molmo 2 model and training documentation](https://github.com/allenai/molmo2/blob/main/README.md)

## Continue exploring

- [Open Models](https://brightaifuture.com/open-models)
- [Open intelligence thread](https://brightaifuture.com/threads/open)