As decision model use skyrockets, is customization the next frontier?

TechCrunch
Read full story· 4 min
Decision models, a new AI type designed for fast, confident choices among preset options, have rapidly gained usage since their October debut. Microsoft launched its own decision model, Microsoft-Decision-1.1, while startups like Sqwish raise funds to offer customizable versions that improve with client data.

What happened

Since October, decision models—a new kind of AI designed to quickly select among preset options—have seen rapid adoption, with over a quarter of Vercel’s AI requests using them just a month after their introduction. Microsoft announced the immediate rollout of its own decision model, Microsoft-Decision-1.1, on Azure. Meanwhile, startups like Sqwish are raising funds to offer customizable decision models that can be improved using client data. The market now features a range of models from major tech companies and startups, each focusing on accuracy, speed, and adaptability for enterprise use.

Key details

  • TypeSafe’s Jev decision model handled over 1.5 billion queries on OpenRouter in one week.
  • Microsoft-Decision-1.1 is now available for wide use on Azure.
  • Sqwish, a U.K. startup, raised $3 million in pre-seed funding for its customizable decision model, kettei fat.
  • Decision models are tuned to make fast, confident choices and register confidence levels, rather than generating text.
  • Developers can easily swap decision models to test performance in specific use cases.

Why it matters

Enterprise users and developers now have more options for integrating fast, cost-efficient AI decision-making into their products and workflows. Customizable models like those from Sqwish allow businesses to improve performance using their own data, potentially reducing errors and increasing relevance for specific tasks. Companies must now evaluate which model best fits their needs, balancing speed, accuracy, and the ability to adapt over time.


More in Machine Learning