Qwen 2.5 is a family of LLM models with specialised versions available for maths and coding and image processing.
https://qwenlm.github.io/blog/qwen2.5/
VL
QWEN 2.5VL is a VLM that is quite capable. It comes in: 3b, 7b, 32b and 72b variants, the latter of which performs on-par or better than GPT-4o