Qwen
Website: https://github.com/QwenLM/Qwen2.5
Options: Local/Offline
Price: Free (downloadable)
The Qwen series of LLMs are free, open-source, and can be downloaded on your own devices. They were created by the internet company Alibaba. There are 3 main models trained for different purposes. One is for text/chat completion, another for coding/programming, and another specifically for math:
- Qwen2.5: 0.5B, 1.5B, 3B, 7B, 14B, 32B, and 72B
- Qwen2.5-Coder: 1.5B, 7B, and 32B on the way
- Qwen2.5-Math: 1.5B, 7B, and 72B.
It also supports over 29 languages, including Chinese, English, French, and Spanish.
Like other “local” LLMs, you can download Qwen and run it on your own computer. However, if your hardware isn’t powerful enough or you prefer an easier method, there are online LLM providers that host these models for you.
Reddit discussions about Qwen
12/02 – What is your favorite model currently?
11/27 – QwQ: “Reflect Deeply on the Boundaries of the Unknown” – Appears to be Qwen w/ Test-Time Scaling
11/21 – M4 Max 128GB running Qwen 72B Q4 MLX at 11tokens/second.
11/15 – Qwen 2.5 7B Added to Livebench, Overtakes Mixtral 8x22B and Claude 3 Haiku
11/14 – Qwen 32B Coder-Ins vs 72B-Ins on the latest Leetcode problems
11/12 – Qwen-2.5-Coder 32B – The AI That’s Revolutionizing Coding! – Real God in a Box?