Incident sheet · /enterprise-ai
- System
- Enterprise AI
- Filed
- Read
- 3 min · 629 words
- Severity
- Routine
Microsoft Ships Decision-Decision Model as Category Hits 25% of Router Traffic
Over a quarter of Vercel router requests now go to decision models. Microsoft ships Decision-1.1 on Azure as Sqwish raises $3M to win on customer-data customization.

Event log
- More than one in four Vercel AI router requests went to decision models in October.
- Microsoft made Microsoft-Decision-1.1 available for wide use on Azure on Tuesday.
- TypeSafe reached a $7.5 billion valuation after its Jev models drew over 1.5 billion OpenRouter queries in one week.
- Sqwish announced a $3 million pre-seed round for its kettei fat decision model.
More than one in four requests on Vercel's AI model router went to decision models in October — a category invented only a month earlier. Microsoft moved to capture that demand on Tuesday. CEO Satya Nadella said Microsoft-Decision-1.1 is available for wide use on Azure immediately.
Decision models are LLMs (large language models) that produce no output tokens; instead they apply general intelligence to choose quickly and cheaply among preset options, registering a confidence level for each choice. In production environments, fast decisions often matter more than the extended reasoning of a slower, costlier frontier model.
Who dominates the category?
TypeSafe set the pace. The company reached a $7.5 billion valuation after releasing its Jev family of models, which continues to dominate usage on services where metrics are visible. OpenRouter alone handled more than 1.5 billion queries to TypeSafe models last week.
Building a decision model is easy, and dozens followed Jev. One Amazon engineer built one in a weekend. Every hyperscaler now fields its own option, and Microsoft this week became the latest to go wide with Decision 1.1.
Where is the startup money going?
Sqwish, a UK-based AI startup behind the decision model kettei fat, announced a $3 million pre-seed round today. Cofounders Ushnish Sengupta and Federica Freddi met as Cambridge graduate students. They had been developing small LLMs when Jev arrived, and realized their architectures could be repurposed as a decision model.
Data is the differentiator, according to the category's leaders. TypeSafe CEO Diogo Almeida said the hardest part of building Jev was finely calibrating its judgment with statistically adjusted synthetic data (artificially generated training examples weighted to mirror real distributions). Sqwish also works with synthetic data, but its core offering is built-in post-training tools that improve model judgment using a customer's own data.
CEO Ushnish Sengupta told TechCrunch: "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."
Do benchmarks matter?
Ranking decision models is difficult. Sqwish's kettei fat has performed well on some benchmarks, but most builders argue the real proof is in production, where developers can substitute one model for another and measure results in context.
"The benchmark is more of an invitation to try the model out, and then the continual tuning is the reason why they should stick to us," Sengupta said.
Microsoft's model also reflects careful balancing across cost, latency, intelligence and calibration, according to Achint Srivastava, the engineering VP who led the project. He warned against over-indexing on benchmarks.
"What we see is that 'benchmaxed' models fall flat under real enterprise usage very quickly," Srivastava wrote in an email. "Our testing underscores just how challenging it is to deliver accuracy, reliability, speed and cost efficiency in a single model. Microsoft-Decision-1.1 and Jev are among the few that perform strongly across all four dimensions, even when evaluated on 'held out' or unseen benchmarks reflective of real use cases."
Can a startup survive the hyperscalers?
For Sqwish, competition with hyperscalers and frontier labs is an opportunity rather than a threat. Freddi predicts future systems will pair System One intelligence — fast, decision-oriented models — with System Two intelligence, the slower reasoning of 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."
via techcrunch.com (Source)
More from Rebecca Stone
Show full bio
Senior reporter covering money, funding and market shifts around enterprise information technology and the leaders who run at CIO Monitor.
199 articles
The newsroom digest
One email a day with the stories that matter. No noise.
Cross-references
- Jev and the Arrival of Model Tetris: Decision Models Get a Standard API
- Microsoft's MAI models chase cheap tokens, not bragging rights
- Decision Models Land in the Enterprise AI Stack at 4 Cents Per Million Tokens
- Microsoft Ships Decision-1, a Cheaper AI Model Built for Verdicts, Not Prose
- Anthropic's $1.4tn IPO Filing Signals Steeper Enterprise AI Bills