# apple/ml-fastvlm

This repository contains the official implementation of "FastVLM: Efficient Vision Encoding for Vision Language Models" - CVPR 2025

Repository: https://github.com/apple/ml-fastvlm
Canonical: https://ross.abutalabs.com/products/ml-fastvlm
Language: Python
License: NOASSERTION
License Family: other
Last push: 2025-05-05T22:59:29+00:00

## Health v2 (maintenance only)
Score: 28/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 20, release rhythm 35, longevity 34
- inputs: {"age_days": 489, "days_push": 485, "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 7411, forks 561 (observed 2026-08-28T04:09:59.747261+00:00)

## What it is
Official implementation of FastVLM, a vision language model with an efficient hybrid vision encoder (FastViTHD) that reduces token count and encoding latency for high-resolution images. Includes pretrained checkpoints (0.5B-7B), inference code, and a demo iOS app for on-device use.

## Use cases
- run a fast vision language model for image question answering
- reduce time-to-first-token for VLM inference on high-resolution images
- deploy a vision language model on iPhone or iPad
- benchmark efficient vision encoders against LLaVA-OneVision
- finetune a VLM with a faster vision encoder
- run multimodal image understanding on mobile devices

## When to choose
- you need low-latency VLM inference, especially on Apple/mobile hardware
- you want pretrained efficient VLM checkpoints with a demo app
- TTFT and vision encoder size are your bottleneck

## When to avoid
- you need text-only LLMs or non-image modalities
- you need a production-supported framework with broad ecosystem tooling
- you require training from scratch rather than LLaVA-based finetuning
- you need non-Apple platform mobile deployment support

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, llm-inference, image-processing, sdk
- domain: machine-learning, deep-learning, computer-vision, large-language-models, mobile-development
- platform: python, cross-platform
- tags: vision-language-model, vision-encoder, multimodal, on-device-inference, cvpr-2025, llava, coreml, ios, macos

## Member repositories
- apple/ml-fastvlm (main) score 28

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:59.747261+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 2026-08-29T17:37:50.275898+00:00, confidence not recorded.
  - readme: https://github.com/apple/ml-fastvlm (fetched 2026-08-28T04:09:59.747261+00:00, sha bd8d07d3302f)
- Data as of 2026-08-30T08:39:29.467469+00:00.