Decision models surge as the industry looks toward custom systems

Executive Briefing
“Computerized decision models are capturing a massive share of traffic as major technology...”
Key Takeaways
- 1.In October, more than one in four requests on Vercel’s AI model router went to decision models rather than large language models.
- 2.These models do not produce output tokens, instead being tuned to quickly choose between preset options while registering confidence levels.
- 3.Microsoft rolled out its Microsoft-Decision-1.1 model for wide use on Azure, while startup Sqwish announced a $3 million pre-seed funding round.
Computerized decision models are capturing a massive share of traffic as major technology firms and startups rapidly deploy new options. Industry builders suggest that continuous fine-tuning on proprietary data will define the next competitive frontier.
Key Developments & Primary Facts
In October, more than one in four requests on Vercel’s AI model router went to decision models rather than large language models.
These models do not produce output tokens, instead being tuned to quickly choose between preset options while registering confidence levels.
Microsoft rolled out its Microsoft-Decision-1.1 model for wide use on Azure, while startup Sqwish announced a $3 million pre-seed funding round.
Strategic Analysis & Implications
Bypassing text generation allows these models to achieve high processing speeds and low costs, making them ideal for production environments.
As hyperscalers flood the market with alternative options, smaller firms are carving out advantages through custom tooling and proprietary customer data.
Outlook & Factors to Watch
The exact performance of individual models across highly specialized enterprise workflows remains to be fully seen beyond initial deployments.







