other
2.25B parameters
Commercial OK
Reviewed May 2026

SmolVLM Instruct

SmolVLM-Instruct is Hugging Face's compact vision-language model built on the Idefics3 architecture, pairing SmolLM2-1.7B-Instruct with a SigLIP-SO400M vision encoder. It is engineered for minimum VRAM footprint and ships under Apache-2.0 with full training-data disclosure (the_cauldron, Docmatix).

License: apache-2.0·Context: 8,192 tokens
BLK · VERDICT

Our verdict

OP · Fredoline Eruo|VERIFIED MAY 29, 2026
unrated

The 'minimum viable VLM' for teams that want a vision model with a clean Apache-2.0 license and a fully published recipe.

Overview

SmolVLM-Instruct is Hugging Face's compact vision-language model built on the Idefics3 architecture, pairing SmolLM2-1.7B-Instruct with a SigLIP-SO400M vision encoder. It is engineered for minimum VRAM footprint and ships under Apache-2.0 with full training-data disclosure (the_cauldron, Docmatix).

Strengths

  • Smallest credible open VLM — runs in ~5GB VRAM at int8
  • Apache-2.0 with published training datasets
  • Strong document understanding from Docmatix training
  • Active follow-ons (SmolVLM2, SmolVLM-256M/500M) show ongoing investment

Weaknesses

  • Behind Qwen2-VL-2B on most public VLM benchmarks
  • Idefics3 image preprocessing is finicky to wire up correctly
  • 8K context limits multi-page document workflows
  • Smaller community than Qwen-VL line means fewer integrations

Quantization variants

Each quantization trades model quality for file size and VRAM. Q4_K_M is the most popular starting point.

QuantizationFile sizeVRAM required
Q4_K_M1.2 GB2 GB

Get the model

HuggingFace

Original weights

huggingface.co/HuggingFaceTB/SmolVLM-Instruct

Source repository — direct quantization required.

Hardware that runs this

Cards with enough VRAM for at least one quantization of SmolVLM Instruct.

Compare alternatives

Models worth comparing

Same parameter band, plus what's one tier above and below — so you can decide what actually fits your hardware.

Frequently asked

What's the minimum VRAM to run SmolVLM Instruct?

2GB of VRAM is enough to run SmolVLM Instruct at the Q4_K_M quantization (file size 1.2 GB). Higher-quality quantizations need more.

Can I use SmolVLM Instruct commercially?

Yes — SmolVLM Instruct ships under the apache-2.0, which permits commercial use. Always read the license text before deployment.

What's the context length of SmolVLM Instruct?

SmolVLM Instruct supports a context window of 8,192 tokens (about 8K).

Source: huggingface.co/HuggingFaceTB/SmolVLM-Instruct

Reviewed by RunLocalAI Editorial. See our editorial policy for how we research and verify model claims.

Related — keep moving

Before you buy

Verify SmolVLM Instruct runs on your specific hardware before committing money.