Jev launches: a System One model for decisions
TypeSafe founder Diogo Almeida introduces Jev, its RLCD training method, and its focus on software decisions. The post includes a 2:56 launch video.

Baseten releases joey-1, a 2B parameter model for classification and scoring without free-text generation. Inference costs just $0.01 per million tokens, with 20ms P50 latency and a 63.30 score on Jev Decision Index 0.4, compared to 72.24 for elephant-1.
⚡ Baseten launches joey-1: AI decisions for USD $0.01 per million tokens The 2 billion parameter model classifies and scores, without generating free text. Baseten reports 20 ms P50 latency and 63.30 points on Jev Decision Index 0.4, compared to 72.24 for elephant-1.