Released weights Qwen 3.6 35B-A3B

The model has been upgraded, so it now often outperforms the Qwen 3.5 27B dense model (I hope the situation with tool calling has been fixed). They likely achieved this by increasing the amount of reasoning again — Qwen models are traditionally very talkative. The main competitor, Gemma 4, although slightly lagging in benchmarks, uses significantly fewer tokens for these results. In the division that deals with Qwen, there are currently major changes focused on profitability, which is why they have been releasing fewer models. But fortunately, we are still getting decent releases. Weights