Back to All Resources

Qwen2.5-7B-Instruct-1M

First open-source Qwen model supporting 1M-token contexts. Features 3-7x faster processing with sparse attention integration and vLLM-based inference framework. Includes technical report detailing training/inference optimizations.