Laya decision model now available on AI Gateway, free through October 31
Source
vercel.com
Author
Zachary Chen
Date
Context
Convai Innovations' Laya evaluates supplied context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition → against questions with defined answer types, returning choices, scores or yes/no answers with probabilities. The AI Gateway announcement illustrates this with a refund request: an application can ask whether the message requests a refund and use the answer to route it.
The trial has an operationally important ending. Through October 31, 2026, the specified promotional provider serves both listed model IDs free. Afterward, the ID ending in “-free” stops serving, while the standard ID begins billing. Choosing between them determines whether the application's trial ends with stopped requests or paid usage.
Key takeaways · AI-distilled
Vercel's suggested uses are concrete: detect a refund request and route it to billing, choose which tool or workflow should handle an AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → request, flag a request or response for policy review, and rate ticket urgency or answer quality on a defined scale.
There are two model IDs. convaiinnovations/laya-free routes to Boundless and stops serving after the October 31, 2026 promotion ends, while convaiinnovations/laya keeps working and begins billing at that point.
Laya is callable through the AI SDK's experimental_evaluate function, a native HTTP API, or a TypeSafe-compatible API. Vercel's example pins the request to the free provider with the gateway option only: ['boundless'].
Terms in this piece · Glossary
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.