Verified cascade
OpenRouter cookbook: Jev checks a cheap model's draft against grounding constraints before accepting it or escalating to a stronger model.
This summary is based on how curated lists describe the project; its README was not available to review.
Overview
This OpenRouter cookbook page describes a verified cascade. According to the list summary, a cheaper model writes a draft, Jev checks that draft against grounding constraints, and the draft is either accepted or the request escalates to a stronger model.
It is a reusable pattern for cutting generation costs while keeping a typed quality gate, and fits naturally with OpenRouter's access to both generative models and Jev.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Does the draft satisfy the grounding constraints?(inferred) | yes · no |
How it works
- A cheap model generates a draft.
- Jev checks the draft against grounding constraints.
- Accept the draft if it passes.
- Otherwise escalate to a stronger model.
Strengths
- Most requests can be served by the cheaper model.
- Typed gate is cheap compared with a frontier judge.
Caveats
- Known only through a list summary.
- Gate quality depends on how constraints are phrased.
Getting started
See https://openrouter.ai/docs/cookbook/evaluate-and-optimize/jev-verified-cascadeAt a glance
- Type
- Article
- Runtime
- OpenRouter API
Use cases and integrations
How curated lists describe it
OpenRouter example that checks a cheaper draft against grounding constraints with Jev before accepting it or escalating to a stronger model.
cobanov/awesome-jev
Frequently asked questions
What does Jev check?
Whether a cheap model's draft meets grounding constraints.
What happens on failure?
The request escalates to a stronger model.
Where is it documented?
In OpenRouter's cookbook.
Related projects
Cost-, cache- and quota-aware LLM router that classifies each turn with a Jev-class model, prices every route, and checks cheap answers.
ProjectprototypePythonZilliz semantic cache for LLM queries, listed for a Jev evaluator that uses Noul checks to decide whether a cached answer can be served.
LibraryproductionPythonLocal-first routing engine for long-running agent tasks that uses a Jev decider to pick a model group per turn.
PluginproductionRust, TypeScriptOpen-source AI gateway for 100+ LLMs that can use Jev to classify requests for complexity routing and to compact tool results.
ProxyproductionPythonOpen-source workspace for running and reviewing AI coding agents across devices, with an optional Jev-based automatic model router.
ProjectproductionTypeScript- jev-router★ 593
Per-turn model routing for Claude Code and Codex: wraps the real CLI and lets Jev send simple turns to fast models and hard ones to strong.
ProxybetaJavaScript