SmolLM
Efficient Mini Language Model
- Supports multiple parameter scales: 135M, 360M, and 1.7B parameters.
- Train on high-quality datasets to provide high performance.
- Suitable for local devices, reducing inference costs and improving privacy protection.
- Excellent performance in various benchmark tests, testing common sense reasoning and world knowledge.
- Supports multiple hardware configurations, from smartphones to laptops.
- Provide ONNX and WebGPU demonstrations that are easy to deploy and use.
- Support instruction tuning, using publicly licensed instruction datasets for training.
Product Details
SmolLM is a series of latest small language models, including versions with 135M, 360M, and 1.7B parameters. These models are trained on carefully planned high-quality training corpora and can run on local devices, significantly reducing inference costs and improving user privacy. The SmolLM model performs well in various benchmark tests, testing common sense reasoning and world knowledge.
