FastVLM 7B

By our estimate, FastVLM 7B needs about 5.9 GB at Q4, 10 GB at Q8 and 18.4 GB at FP16 of VRAM. That covers the weights plus headroom for a short context, and longer contexts need more.

Yes, by our estimates FastVLM 7B fits on an 8 GB GPU at Q4 (about 5.9 GB) and at higher precision on larger cards.

Yes, a Mac with at least 8 GB of unified memory should run FastVLM 7B at Q4, which needs an estimated 5.9 GB. This assumes macOS gives the GPU about 75% of unified memory.

Yes, FastVLM 7B has open weights, so you can run it on your own hardware and keep prompts on your own machines.

Run FastVLM 7B locally

Yes, FastVLM 7B accepts images as input alongside text.

Similar models from other labs include Llama Prompt Guard 2 22M (Meta), Llama Prompt Guard 2 86M (Meta) and Qwen3 8B (Alibaba). They are the closest matches by kind and release date.

Llama Prompt Guard 2 22MFastVLM 7B vs Llama Prompt Guard 2 22MLlama Prompt Guard 2 86MFastVLM 7B vs Llama Prompt Guard 2 86MQwen3 8BFastVLM 7B vs Qwen3 8B

FastVLM 7B has open weights under the apple-amlr license, so you can download and run it yourself.

The FastVLM 7B weights are on Hugging Face at apple/FastVLM-7B.

Run FastVLM 7B locally

FastVLM 7B was made by Apple. It was released on May 7, 2025.

More from Apple

Yes, FastVLM 7B has open weights, so you can run it on your own hardware without API fees.