Skip to content

ON-DEVICE VLM HUB

Small vision-language models that run on-device

A vision-language model answers questions about images. The models below are small enough for laptops, phones or WebGPU browsers. Use the table to shortlist, then open a head-to-head comparison with FastVLM.

ModelDeveloperSizesImageVideoMulti-imageBrowserMLXLicenseBest for
FastVLM2025Apple0.5B · 1.5B · 7BYesApple AMLRLow time-to-first-token on high-resolution images; browser and Apple Silicon
SmolVLM22025Hugging Face256M · 500M · 2.2BSome sizesApache 2.0Tiny models with video understanding; permissive license
LFM2.5-VL2026Liquid AI450M · 3BYesLFM Open LicenseGrounding, screens, documents and tool calling at the edge
Gemma 42026GoogleE2B · E4B · 12B · 26B-A4B · 31BRuntime-dependentGemma TermsBroad multimodal (image, video frames, audio on E2B/E4B) and agentic use
Gemma 3n2025GoogleE2B · E4BRuntime-dependentGemma TermsPhone-first models with audio input and per-layer embeddings
Qwen3-VL2025Alibaba2B · 4B · 8B (small)Some sizesApache 2.0Strong OCR, documents and video in small dense sizes
Moondream2024–2025Moondream2B · 3 (9B MoE, 2B active)Some sizesApache 2.0 / BSLPointing, detection and captioning with a tiny footprint
MiniCPM-V2025OpenBMB4.5 (8B)NoApache 2.0 code · MiniCPM weightsHigh-resolution image and video understanding on laptops
PaliGemma 22024Google3B · 10B · 28BNoGemma TermsFine-tuning base for detection, segmentation and captioning

“Browser” means a maintained WebGPU/Transformers.js or ONNX build exists for at least one size. “Runtime-dependent” models run through vendor runtimes such as MediaPipe or LiteRT rather than Transformers.js.

Facts checked on 2026-09-17 against official model cards and announcements. Report an outdated row via the site’s GitHub.

How to choose

  1. Start from the task: OCR and documents favour higher input resolution; video needs a model trained on frames; grounding needs boxes in the output format.
  2. Then the runtime: browser demos need ONNX weights; iPhone apps need MLX or Core ML; Linux servers can use anything.
  3. Then the license: Apache 2.0 models are simplest to ship; Apple AMLR and Gemma terms have conditions you must read.
  4. Finally measure on your device—published benchmarks use different protocols and cannot be compared across vendors.