Technology news and discussion on TalkFuse.
Decision models gain traction as firms explore customer-specific training
Decision models are gaining traction as businesses seek AI that can make quick, low-cost choices among preset options, with a confidence estimate. Their spread matters because they may handle routine production decisions more efficiently than slower, more expensive language models, while competition grows among established technology firms and startups.
Vercel reported that more than a quarter of requests through its model router in October went to decision models. TypeSafe’s Jev models remain widely used, while Microsoft has launched Microsoft-Decision-1.1 on Azure and UK startup Sqwish has raised $3 million in pre-seed funding. Sqwish’s kettei fat model is designed to improve using customers’ own production data; companies say real-world accuracy, reliability, speed and cost matter more than benchmark rankings alone.
In short
- Decision models are taking a growing share of AI requests.
- Microsoft and startups are entering the fast-moving market.
- Sqwish offers customer-specific tuning using production data.
Explained simply
AI models do not all generate open-ended text. Decision models choose between a set of possible answers and can estimate how confident they are, making them suited to tasks such as sorting requests or routing work.
Businesses may use them for routine choices where speed and cost matter. The field includes established technology companies and newer firms, with some models designed to learn from a customer’s own production data.
Their usefulness depends on how accurately and reliably they perform in real settings, as well as how quickly and cheaply they can do the job. That makes customer testing an important part of judging them alongside standard benchmark results.
Read the full article at the source →
Originally published by TechCrunch as “As decision model use skyrockets, is customization the next frontier?”.