Jev 1.13 (AI model): news timeline & CVEs

Liquid AI Releases d1: A Decision Model That Returns Calibrated Probabilities With Zero Output Tokens

Liquid AI launched d1, a hosted decision model returning calibrated probabilities with zero output tokens.

Liquid AI released d1, a hosted decision model that answers typed questions with calibrated probabilities and zero generated tokens. Its primitives are Noul for yes/no probabilities, Choice for a labeled option distribution, and Score for rubric positions, all evaluated in one call to the Liquid API as d1:free. Liquid says there are no downloadable weights and the model is not trainable. The company contrasts it with generative LLMs and with decision models such as TypeSafe Jev 1.13, Convai Laya, and JEV-27B.

MarkTechPostSep 29, 21:47 UTC · 1d agoModel release

20 Agentic Use Cases of TypeSafe AI’s Jev

TypeSafe AI launched Jev, a closed decision model that returns typed, calibrated choices for agent loops.

TypeSafe AI released Jev 1.13, a closed hosted System One model that returns typed Choice, Score, and Noul decisions with probabilities calibrated by reinforcement learning. Founder Diogo Almeida previously worked at OpenAI on instruction-following research. Vendor workflow evals claim up to 193.6x speed and 444.6x lower cost versus GPT-6 Astra and Fable 5.1, with roughly 70–500 ms latency and $0.042 per million input tokens. The post lists 20 agent uses, including model routing, tool-call risk gating, prompt-injection screening, reranking, and browser or desktop control, and compares Jev with open models Laya and kev plus Claude Opus 5.

MarkTechPostupdated Sep 30, 19:00 UTC · 9h agofirst Sep 28, 03:07 UTC · 3d agoModel release 3 sources2

Jev 1.13

Timeline2 articles

Appears with

Liquid AI Releases d1: A Decision Model That Returns Calibrated Probabilities With Zero Output Tokens

20 Agentic Use Cases of TypeSafe AI’s Jev