Skip to content
United States of America
File

Baseten: AI decision models

This dossier tracks Baseten decision models, availability, pricing and evaluation.

Last updated : 1 key moments

Where things stand

As of 29 October 2026, joey-1 is announced on API and as open weights; speed results remain vendor-reported.

Baseten launches joey-1 for low-cost classification and routing

Baseten announced joey-1 on 29 October 2026, a two-billion-parameter decision model derived from Qwen3.5-2B. It is available through its API and as open weights, at an announced API price of $0.01 per million tokens.

Rather than write free-form answers, it returns a label, decision or score with a probability. Target uses include support-request routing and automated checks.

Baseten reports median latency of 20 milliseconds for an average 2,000-token request. That measurement depends on its test configuration and does not establish latency for every deployment. Users must assess accuracy on their own data before delegating sensitive decisions.

Sources 3