Sakana Chat updated: new model refresh, memory feature added

Sakana AI has updated Sakana Chat with two changes that matter more together than apart: the service now uses Fugu Max, and it can remember what you told it in earlier conversations.

Fugu Max is the orchestrator model Sakana AI recently exposed through its API, and now it’s available to every Sakana Chat user. The idea is simple enough, even if the machinery behind it is not: it pulls from a mix of open models and sends each prompt to whichever one fits best, instead of leaning on a single model for everything. Sakana says that approach can match frontier-model performance, and now anyone using the model picker can try it directly.

Sakana Chat itself has been around since March 2026, when Sakana AI launched it alongside the alpha version of Sakana Namazu as a place to test its models without going through an API. In August, the company added Sakana Fugu, model selection, code execution, and support for image and document attachments. This latest refresh pushes that setup a bit further by swapping in Fugu Max as the default Fugu option.

The new memory feature may be the more practical change. Before this update, every conversation started fresh. Now, things like your role, your project, and your preferred style can carry over into later chats, so you don’t have to repeat yourself each time. Memory works with both Namazu and Fugu Max, can be reviewed from settings, and can be turned off. Sakana AI says it only applies to conversations from this update onward, not older ones.