TypeSafe AI
Jev
A hosted decision model for choosing labels, rating a rubric and estimating yes/no probabilities.
Reviewed deploymentHostedCheck serving requirements
API price$0.042 / 1M input tokensInput usage determines cost
jev-1.13.0
MODELS / DECISION & CLASSIFICATION
Compare decision and classification models.
Choose a label. Check a condition. Rate a clear rubric. Find a model that fits your language, hardware and workflow.
START WITH YOUR CONSTRAINTS
API convenience or local control?
Compare the access you need first. Then test accuracy, uncertainty, latency and cost on the same decisions.
Jev vs Laya vs Kev: the selection guide →Find your deployment fitSources checked 2026-10-05
Filters use documented capabilities. Unverified support is excluded when a requirement is selected. Language quality and deployment still need your own checks.
6 modelsEditorial order · no performance ranking
Convai Innovations
Laya
An open-weight encoder family with English and multilingual checkpoints for short, typed decisions.
Reviewed deploymentSelf-hostedCPU support documented
API priceHosted price unverifiedSelf-hosting has operating costs
Laya family · model card checked 2026-10-05
Jared Palmer
Kev-4B
An open decision model with a pointer head, typed probabilities and a Jev-compatible serving interface.
Reviewed deploymentSelf-hostedCheck serving requirements
API priceHosted price unverifiedSelf-hosting has operating costs
Kev 1.0 · Kev-4B
Fastino
GLiNER2.5-Decide
A compact open-weight model combining schema-defined decisions with extraction and constraints across outputs.
Reviewed deploymentSelf-hostedCPU support documented
API priceHosted price unverifiedSelf-hosting has operating costs
GLiNER2.5-Decide · 2026-09-24 release
Bespoke Labs
Bespoke Nimble
A decision model and public training recipe for teams exploring their own data, evaluations and deployment.
Reviewed deploymentSelf-hostedCheck serving requirements
API priceHosted price unverifiedSelf-hosting has operating costs
Bespoke-Nimble-9B · 2026-09-24 checkpoint update
Together AI
Tev1
An experimental fine-tuned model for selecting one supplied option, with a public classifier-training recipe.
Reviewed deploymentHosted / Self-hostedCheck serving requirements
API priceHosted price unverifiedSelf-hosting has operating costs
Tev1-4B-experimental · new-v1 recipe
Select up to three models to compare. A probability, a confidence statistic and a token logprob are not interchangeable. Test the whole task before setting thresholds.
PUT A DECISION TO WORK
Learn the pattern. Measure the result.
Route an agent request with a decision model
Define clear intent labels, keep uncertain cases visible and validate the selected tool before it runs.
Check RAG evidence before generating an answer
Evaluate relevance and evidence coverage after retrieval, and measure what filtering removes.
When do decision models actually save money?
Measure complete-task cost, fallback frequency and accepted quality before claiming token savings.