SmolLM3
SmolLM3
Three billion parameters, fully open, and it will run on a machine you already own.
Updated
The read
SmolLM3 is Hugging Face's own small model, and it is open in the fullest sense: weights, data mix and training recipe all published under Apache 2.0. It has a dual-mode setup that lets you turn reasoning on when a task needs it, handles a long context for its size, and runs on hardware down to a hobbyist board. It is a three-billion-parameter model, so calibrate accordingly.
Where it shines
- Weights, data and training recipe all published
- Runs on very modest hardware, down to a single-board computer
- Reasoning mode you can switch on only when it is worth it
Worth knowing first
- Capability ceiling is low, and it is reached quickly
- Supports far fewer languages than the bigger open families
Index score
67.7
Provisional. This score is read from the product's public capability, not from the one-prompt rebuild the fully reviewed entries went through, so treat it as a placing rather than a verdict.
- Rank
- 73 of 73
- Builds
- LLMs
- Category
- Small and fast
- Output
- Apache-licensed weights, data and recipe
- Pricing
- Free tier
- Best for
- Local and hobbyist deployment with nothing hidden
Scores are our own editorial judgement, weighted craft 55, speed 10, control 15 and value 20.
Also worth a look
Full index- 11Claude Haiku 5.5Claude Haiku 5.5ProvisionalA small model priced like a rounding error that scores where mid-tier models did last quarter.Small and fastPaid only87.9
- 22GPT-6 LunaGPT-6 LunaProvisionalThe cheap end of GPT-6, and the model free ChatGPT users now talk to.Small and fastPaid only84.9
- 30GPT-5.6 LunaGPT-5.6 LunaProvisionalStill cheap and still capable, but GPT-6 Luna costs half as much.Small and fastPaid only83.0
