Qwen3.8 2.4T A95B is an open-weight sparse mixture-of-experts model from Qwen, the open-weight variant of Qwen3.8 Max, with 95 billion active parameters out of 2.4 trillion total and a 1,010,000-token context window.
