Qwen3-1

Qwen3-1.7B is a highly efficient 1.7-billion-parameter open-weight model featuring a native thinking mode for advanced reasoning and a 32K context window.

Qwen3-1.7B packs advanced reasoning into a compact 1.7-billion-parameter footprint, making it ideal for local deployment on standard consumer hardware. Built by Alibaba Cloud, the model features a unique hybrid architecture that lets users switch seamlessly between a high-speed, general-purpose non-thinking mode and a deep-reasoning thinking mode (utilizing reinforcement learning and chain-of-thought processing). It supports a native 32K-token context length, handles tool integration with precision, and delivers competitive performance on math, coding, and multilingual tasks across 100-plus languages.

https://github.com/QwenLM/Qwen3