Skip to main contentThe Local Bench

Vision-language · Hugging Face

SmolVLM2 2.2B

Hugging Face: HuggingFaceTB/SmolVLM2-2.2B-Instruct · checked 2026-10-13

Tiny image and video model for phones, Raspberry Pi-class boards and browsers.

Datasheet

Maker
Hugging Face
Parameters
2.25 BVerifiedsafetensors.total from the HF API Source
Licence
Apache-2.0Verified
Access
Public download
Task (HF pipeline)
image-text-to-text
First published on HF
2025-02-08
Last updated on HF
2025-04-08
Weights in memory
4.5 GB at 16-bit · 1.4 GB at ~4.85 bitsEstimateParameters × bits ÷ 8. Excludes context cache and runtime overhead; not every model has a 4-bit build.

Benchmark

42%
MMMU (val) · #3 of 3 scored in this category

MMMU column, model card table. Source

Other published results

See the full vision-language ranking

Used in

The description summarises the model card at https://huggingface.co/HuggingFaceTB/SmolVLM2-2.2B-Instruct. Parameters and licence are read from the Hugging Face API; the weights-in-memory line is our arithmetic, not a measurement.