5 points | by zyruh 2 days ago ago
3 comments
Allen AI's OLMo2. It is the only truly open-source model I've heard of.
The training code and dataset are also open-sourced.
GLM-4.5 from https://chat.z.ai because it renders math notation so well that other LLMs tend to mangle and it uses color coding when showing Python code.
Qwen3 30B 2507 is a good mix of speed and performance, for my taste. I use an abliterated quant and have no complaints.
Allen AI's OLMo2. It is the only truly open-source model I've heard of.
The training code and dataset are also open-sourced.
GLM-4.5 from https://chat.z.ai because it renders math notation so well that other LLMs tend to mangle and it uses color coding when showing Python code.
Qwen3 30B 2507 is a good mix of speed and performance, for my taste. I use an abliterated quant and have no complaints.