Description

OpenJev is a "System One" decision server that provides calibrated probabilities and confidence scores for typed questions by reading token probabilities directly from the model's canvas. It runs on DiffusionGemma via vLLM, supports images, and maintains compatibility with the Jev API while preventing hallucinations through its non-generative inference method.

Descriptions, tags, and model credits may be AI-generated or inferred from public sources and can be incomplete or wrong. Learn more.