Laya AI

SHORT TEXT · TYPED ANSWERS

Laya AI for the decisions between data and action.

Turn a message into a yes/no probability, a category, or a score. Try Laya AI in the browser, then send the same request from your application.

English and multilingual models. The playground checks live availability before a run.

WORKSPACE

Make a decision from a short text

Write the input and the rule, then run an available Laya model to get a typed answer.

512 total tokens per question, including text, instruction, and labels. Oversized requests are rejected.
105 characters; token budget is checked by the model gateway.
Separate labels with commas.
1 credit per successful run

Checking model availability…

STRUCTURED RESULT
Answers appear here

Choose a model and submit a short decision. No sample result is presented as a live response.

ONE MODEL, FOCUSED JOBS

Where Laya AI fits

Small decisions are easier to ship when the answer has a predictable shape.

01

Sort incoming messages

Choose a support queue from a short ticket and keep the distribution for review.

02

Check a policy rule

Ask one yes or no question about a comment before an automated action.

03

Score a compact record

Return a level on a rubric you define, with the underlying probabilities.

THE CONTRACT

Ask what you need. Handle what comes back.

Laya reads short text with explicit typed questions. The model does not write long-form prose. Oversized requests are rejected before silent truncation can hide missing context.

How Laya works →

NEXT STEPS

Build with a clear model choice

Use the source and model cards for local setup, or compare the same question across decision models.

COMMON QUESTIONS

Before you build

What is Laya AI?

Laya AI is a playground and API for short-text decisions using the Laya models. It returns typed answers rather than chat prose.

Which Laya model can I choose?

Choose the English or multilingual variant. The playground shows current service availability.

What happens to long input?

The service rejects input that exceeds the model token budget rather than silently truncating it.