Early this morning, Hugging Face, a world-famous open platform for large models, opened the open source, the top small-parameter model SmollM3. SmollM3 has only 3 billion parameters, yet its performance greatly exceeds similar open source models such as Llama-3.2-3 B and Qwen2.5-3B. It has a 128k context window and supports 6 languages including English, French, Spanish, and German. It supports deep thinking and non-thinking dual reasoning modes, and users can switch flexibly.

Zhitongcaijing · 07/08 22:57
Early this morning, Hugging Face, a world-famous open platform for large models, opened the open source, the top small-parameter model SmollM3. SmollM3 has only 3 billion parameters, yet its performance greatly exceeds similar open source models such as Llama-3.2-3 B and Qwen2.5-3B. It has a 128k context window and supports 6 languages including English, French, Spanish, and German. It supports deep thinking and non-thinking dual reasoning modes, and users can switch flexibly.