This is a continuous finetuned vesion of Qwen2.5-72B-Instruct, with higher quality data.

Compared to previous models, we modified the data and make their prompts' diversity better

As a result, it has a superb ability in instruction following. On the other side, we find its coding ability drop a little bit, guessing as a result of descriptions not formal enough.

Downloads last month: -

Safetensors

Model size

73B params

Tensor type

BF16

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support