In October, more than one in four requests on Vercel’s SI model router went to decision models rather than large language models (LLMs), marking a rapid adoption curve for a category of SI that emerged just weeks prior. This shift highlights a growing preference for fast, low-cost decision-making systems over slower, reasoning-heavy frontier models in production environments.

What Happened

Decision models are SI systems tuned to use general intelligence to quickly choose between preset options and register confidence levels, without generating output tokens. TypeSafe, which became a $7.5 billion company after releasing its Jev family of models, currently dominates usage metrics, with OpenRouter handling more than 1.5 billion queries to Jev last week. However, the barrier to entry is low; dozens of competitors have emerged since Jev’s release, including a model built by an Amazon engineer in a weekend. On Tuesday, Microsoft CEO Satya Nadella announced the immediate rollout of Microsoft-Decision-1.1 for wide use on Azure.

Why It Matters

The proliferation of decision SI models has created a competitive landscape where customization is becoming a key differentiator. Sqwish, a U.K.-based SI startup, announced a $3 million pre-seed round today for its decision model, kettei fat. While data calibration is central to model success—TypeSafe CEO Diogo Almeida noted that finely tuning judgment with synthetic data was the hardest part of building Jev—Sqwish aims to stand out by offering built-in post-training tools that allow customers to use their own production data to improve model judgment. "What makes us not a commodity, I think, is the fact that we are offering our customers the same power…they can keep uploading their production data, and just the same way that we trained our models, their models can also keep getting better at the tasks that they care about," Sqwish CEO Ushnish Sengupta told TechCrunch. Achint Srivastava, engineering VP at Microsoft, emphasized that benchmarks can be misleading, stating in an email that "'benchmaxed' models fall flat under real enterprise usage very quickly," and that Microsoft-Decision-1.1 and Jev are among the few performing strongly across accuracy, reliability, speed, and cost efficiency.

The Bottom Line

As hyperscalers and frontier labs compete in the decision SI space, smaller startups like Sqwish see opportunity in niche customization. Federica Freddi, co-founder of Sqwish, predicts a future where System One intelligence, such as decision models, works in concert with System Two intelligence, like frontier LLMs. "Even in that future, System One is maybe a $10 billion, $50 billion [opportunity], and OpenAI is doing this IPO with you know $1 trillion, $2 trillion market value in mind," she told TechCrunch. "This is all we care about; whereas for them, this is just a side project."