AI Compass / Quick reads
A model that only decides costs almost nothing
Routine questions in agent workflows get much cheaper and faster when a decision model answers them and the large language model only writes.
In depth: Decision models – when a model should only choose
Many agent steps are decisions rather than writing tasks: yes/no, a choice from a list, or a score. For these there are now decision models such as Jev from TypeSafe. They return one of the defined answers plus a probability, but they write no text and call no tools. According to OpenRouter, Jev costs $0.042 per million input tokens, and output is free (checked externally).
Tool and alternative
Jev is proprietary and reachable through a dedicated decisions interface at OpenRouter, where the interface is called “alpha”. According to AI LABS (28 September), new sign-ups at TypeSafe itself are paused. Open option: a small local classification model. Or a small language model such as Claude Haiku with fixed answer options: more expensive (compare list prices before choosing), but tied to no single interface.
Sources
- Jack Roberts, “Jev AI Just Dropped”, 19 September 2026 (chapter 10:59 “when Jev, when a large model”)
- OpenRouter model page Jev 1.13, published 18 September 2026, retrieved 2 October 2026
- OpenRouter blog “What Is Jev?”, 21 September 2026
Context: 7 videos on 3 channels in two weeks, some with affiliate links, so not independent enthusiasm. Cost comparisons by the creators: their own measurements, prompts not disclosed.
Want to try this with your team? That’s what the AI Lab is for. See the AI Lab