#long context LLM09/08/2025
Tiny Titans: Qwen3-4B-Instruct-2507 and Qwen3-4B-Thinking-2507 Bring 256K Context to Consumer Hardware
'Alibaba introduces two 4B-parameter models with native 256K token context windows: an Instruct model for concise multilingual responses and a Thinking model for explicit chain-of-thought reasoning, both optimized for consumer deployments.'