QwQ 32B

QwQ 32B is a reasoning-focused, autoregressive large language model: it produces text but is tuned to spend more effort working through problems such as mathematics and coding. QwQ belongs to Alibaba Cloud’s Qwen family; it is not a separate organization or a Meta model. The name is easiest to remember as Qwen’s reasoning branch, while 32B indicates its approximate parameter scale. TT-Metal lists QwQ with tensor parallelism across eight devices (TP=8) on a Wormhole QuietBox and a batch size of 32. Tensor parallelism divides one model across chips, making this a multi-chip deployment without by itself saying anything about the number of host computers.