CorticorpNews
Back to Science & Tech

Updated

Sqwish Labs, a UK AI startup, is pushing customization of decision models by training them on customers' own production data, TechCrunch reported on November 17, 2026. Decision models are LLMs tuned to quickly and cheaply choose between preset options while reporting their confidence, and they are useful where fast decisions matter more than slower frontier-model reasoning. The company, behind a model called kettei fat, announced a $3 million pre-seed round, and its cofounders Ushnish Sengupta and Federica Freddi had repurposed small LLM architectures into decision models after rival TypeSafe's Jev family arrived. Sengupta told TechCrunch that Sqwish's built-in post-training tools let customers keep uploading production data so their models keep improving on the tasks they care about. TechCrunch reported that benchmarks are only a rough guide, since developers can easily swap one decision model for another and test it in their own context. Sqwish competes with hyperscalers such as Microsoft, which is rolling out its own Microsoft-Decision-1.1 model on Azure, and with frontier labs.

Sources:TechCrunch

1 outlet · 1 article · first seen Nov 17, 1:11 PM ET

Every source article