Qwen/Qwen3-4B-Instruct-2507
🧠 AI ModelQwen
4B parameter instruct model from Qwen, top performance on reasoning and multilingual tasks.
Qwen3-4B-Instruct-2507 is part of the Qwen3 series, trained with advanced reinforcement learning techniques to enhance instruction following, reasoning, and safety. It features Grouped Query Attention (GQA) for efficient inference, supports up to 32k token context length, and includes built-in support for tool calling and multi-turn conversations. Despite its compact 4B parameter size, it achieves results competitive with 7B-parameter models on major benchmarks like MMLU, GSM8K, and HumanEval. The model is fully open-source under Apache 2.0 license, making it accessible for both research and commercial use. It is optimized for deployment on consumer-grade hardware and can be run via Transformers, vLLM, or Hugging Face Inference Endpoints.
💡Highlights
- ├─4B params, SOTA on small models
- ├─Apache 2.0, fully open
- └─Tool use & extended context
🎯For
- ├─Developers
- ├─Researchers
- └─AI Enthusiasts