This model is exactly the same as Qwen2.5 1.5B Instruct, but the vocabulary is padded to the same size as larger Qwen models (like Qwen2.5 72B Instruct). This allows it to be used as a draft model in speculative decoding.
Chat template
Files info
