357
Productivity & Workflow354
Software Development248
Automation & Workflow238
Marketing & Growth204
AI Infrastructure & MLOps197
Photography & Imaging166
Writing & Content Creation193
Design & Creative170
Data & Analytics165
Customer Support156
Education & Learning144
Voice & Speech130
Sales & Outreach132
Operations & Admin119
Decision models are taking a growing share of AI traffic. Sqwish raised $3M and says companies can improve these models using their own data.
In short: Decision models, a newer type of AI built for fast yes or no choices, are being used more often, and startups like Sqwish are trying to stand out by letting customers tune them with their own data.
Decision models are gaining traction quickly. TechCrunch reports that in October, more than one in four requests on Vercel’s AI model router went to decision models instead of chat-style AI tools.
A decision model is like an AI multiple-choice picker. It does not write long answers. Instead, it chooses between preset options and says how confident it is. This can be useful when a business needs fast, low-cost decisions, like picking which support ticket should be routed to which team.
TypeSafe’s Jev models helped kick off this wave, and TechCrunch says OpenRouter handled more than 1.5 billion queries to TypeSafe’s models last week. Since then, many competitors have launched similar models. Microsoft also said it is rolling out Microsoft-Decision-1.1 for wide use on Azure.
Sqwish, a U.K. startup, announced a $3 million pre-seed funding round and is promoting its decision model called “kettei fat.” The company says its main pitch is customization after the model is trained, using a customer’s own production data (real data from day to day use) so the model can get better at the specific tasks that customer cares about.
As more companies release similar decision models, the key difference may shift from general benchmark scores to how well a model can be tuned for a specific workplace. That could make tools for safe, ongoing customization more important than simply picking the top model on a public leaderboard.
Source: TechCrunch AI