# apple/ml-aim

This repository provides the code and model checkpoints for AIMv1 and AIMv2 research projects.

Repository: https://github.com/apple/ml-aim
Canonical: https://ross.abutalabs.com/products/ml-aim
Language: Python
License: NOASSERTION
License Family: other
Topics: jax, mlx, pytorch, large-scale-vision-models
Last push: 2025-08-04T17:58:52+00:00

## Health v2 (maintenance only)
Score: 42/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 35, release rhythm 35, longevity 68
- inputs: {"age_days": 964, "days_push": 394, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1424, forks 73 (observed 2026-08-28T04:04:41.367314+00:00)

## What it is
Apple's official repository for AIM (Autoregressive Image Models), providing code and pretrained checkpoints for AIMv1 and AIMv2 large vision encoders trained with a multimodal autoregressive objective. Implementations are available in PyTorch, JAX, and MLX, with models hosted on Hugging Face.

## Use cases
- pretrain or fine-tune large vision encoders with autoregressive objectives
- extract image features from a frozen AIMv2 backbone for downstream tasks
- run open-vocabulary object detection with AIMv2 encoders
- benchmark against CLIP, SigLIP, and DINOv2 on multimodal understanding tasks
- load AIM checkpoints in PyTorch, JAX, or MLX

## When to choose
- you need state-of-the-art frozen vision backbones for classification or detection
- you want to reproduce or build on AIMv1/AIMv2 research
- you prefer Apple-published checkpoints with multi-framework support

## When to avoid
- you need a production inference server rather than research code
- you lack GPU resources for large vision models
- you need a permissively licensed library - the license is custom/non-standard

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning
- domain: machine-learning, computer-vision, image-processing, artificial-intelligence
- platform: python
- tags: vision-encoders, autoregressive-pretraining, model-checkpoints, pytorch, jax, mlx, research-code, gpu

## Member repositories
- apple/ml-aim (main) score 42

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:41.367314+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 2026-08-30T04:37:34.315389+00:00, confidence not recorded.
  - readme: https://github.com/apple/ml-aim (fetched 2026-08-28T04:04:41.367314+00:00, sha 6e6987f50a47)
- Data as of 2026-08-30T08:39:29.467469+00:00.