newsfilter
← Latest News
TechCrunch·

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

Summary

  • In October, over 25% of requests on Vercel's AI model router were directed to decision models rather than LLMs.
  • Decision models are LLMs tuned to quickly select from preset options with confidence levels without generating output tokens.
  • TypeSafe became a $7.5 billion company after releasing its Jev family of models, which handled over 1.5 billion queries last week via services like OpenRouter.
  • Dozens of decision models have been released since Jev, including options from every hyperscaler and Microsoft's new Microsoft-Decision-1.1 model rolling out on Azure.
  • UK-based startup Sqwish announced a $3 million pre-seed round for its kettei fat decision model.
  • TypeSafe CEO Diogo Almedia stated that calibrating judgement using statistically-adjusted synthetic data was the hardest part of building Jev.
  • Sqwish offers built-in post-training tools that use customer data to improve model judgement.
  • Sqwish CEO Ushnish Sengupta noted that customers can upload production data to keep their models improving at specific tasks.
  • Microsoft engineering VP Achint Srivastava stated that "benchmaxed" models often fail under real enterprise usage.
  • Microsoft-Decision-1.1 and Jev are among the few models performing strongly across accuracy, reliability, speed, & cost efficiency on unseen benchmarks.
  • Sqwish cofounder Federica Freddi predicts future systems will combine System One intelligence (decision models) with System Two intelligence (frontier LLMs).
  • Freddi estimates the decision model market opportunity at $10 billion to $50 billion, contrasting it with the trillion-dollar valuations of frontier LLM companies.

Sqwish Labs has a new wrinkle on the decision model: Train it with your own data to max its capabilities.

Keywords
AI,Microsoft,Exclusive