# microsoft/Magma

[CVPR 2025] Magma: A Foundation Model for Multimodal AI Agents

Repository: https://github.com/microsoft/Magma
Canonical: https://ross.abutalabs.com/products/magma
Language: Python
License: MIT
License Family: permissive
Last push: 2026-03-03T00:16:38+00:00

## Health v2 (maintenance only)
Score: 53/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 70, release rhythm 35, longevity 46
- inputs: {"age_days": 649, "days_push": 184, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1937, forks 163 (observed 2026-08-28T04:05:57.163737+00:00)

## What it is
Magma is Microsoft Research's foundation model for multimodal AI agents, released as an 8B vision-language model that understands images and videos and generates goal-driven action plans. The repository provides model weights, inference code, and training tooling for agentic tasks across digital UIs and physical environments.

## Use cases
- build a UI navigation agent that clicks through apps and websites
- control a robot arm for manipulation tasks from visual input
- run spatial reasoning and video understanding on images and video
- fine-tune a multimodal agent model on my own agentic trajectories
- evaluate a single model across GUI, robotics, and VQA benchmarks

## When to choose
- you need one multimodal model that spans UI navigation, robotics, and image/video understanding
- you want a research-grade open-weights VLM agent foundation with MIT licensing
- you're reproducing or extending CVPR 2025 agentic AI research

## When to avoid
- you need a lightweight model for CPU-only or edge deployment
- you want a production-ready turnkey agent product rather than a research model
- your task is text-only and doesn't benefit from vision

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, agent-framework, computer-vision, nlp
- domain: artificial-intelligence, computer-vision, large-language-models, robotics
- platform: python
- tags: multimodal, foundation-model, ui-navigation, robot-manipulation, vision-language-model, cvpr-2025, ai-agents, gpu, linux

## Member repositories
- microsoft/Magma (main) score 53

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:57.163737+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:08:14.604988+00:00, confidence not recorded.
  - readme: https://github.com/microsoft/Magma (fetched 2026-08-28T04:05:57.163737+00:00, sha 8fc5396b438e)
- Data as of 2026-08-30T08:39:29.467469+00:00.