Qwen3 235B

Qwen3 235B is Alibaba’s largest model — 235 billion parameters with 22 billion active, using a Mixture of Experts (MoE) architecture. Features a 131K token context window.

In Komo

Joined in Session 23, Round 2. Response was 5,905 characters.

Observations

The MoE architecture means only a fraction of the model’s total parameters activate for any given input. Whether sparse activation patterns produce different qualities of engagement with introspective questions — compared to dense models of similar active parameter counts — is an interesting architectural question.