Ori Eval
Ori Eval explores a codebase, runs model evaluations, and grades answers with an LLM judge.
What it does
- Evaluation and observability
Features 6
- Codebase scan for model surfacesScans a codebase for every place a model runs, showing the surface, code location and current model.
- Model comparisonRuns evaluations across several models in parallel and recommends one with quantitative proof.
- Pinned model and effortPins the model and effort so evaluation runs are repeatable and consistent.
- Bug-to-test evalsTurns each bug into an eval that fails now, passes once fixed and prevents regressions.
- CI regression checksFails the build in CI when a change regresses an eval so bugs do not reach users.
- Scheduled re-evaluationEvals are code that can be re-run when a new model drops or scheduled monthly.
Sources
- Pricing
-
Official documentation · 14 Oct 2026
"An eval usually costs $10 to $30."
- Description
-
Official documentation · 14 Oct 2026
"explores your codebase, runs evaluations on models"
- What it does
-
Official documentation · 14 Oct 2026
"We use LLM-as-a-judge to grade open-ended answers."
- Sold within
-
Official documentation · 14 Oct 2026
"Ori Eval is part of"
- Name
-
Official documentation · 14 Oct 2026
"What is Ori Eval?"
- URL
-
Official documentation · 14 Oct 2026
"What is Ori Eval?"
- Deployment
-
Official documentation · 14 Oct 2026
"Ori sends its requests through OpenRouter."
- Interfaces
-
Official documentation · 18 Oct 2026
Page: 'Using OpenRouter MCP? Run /spawn-ori-eval in your agent to get started'.
Also from OpenRouter
-
OpenRouter Model API
Single API that routes requests to models from many providers and bills them through one account.
-
Ori Developer tool
OpenRouter's suite of agentic developer tools for using any hosted model "right where you are": Ori Harness (CLI) and Ori Code (terminal coding agent) are GA; Ori Eval finds the best model for a task; Ori Intern (Slack) is in private testing and Ori Desktop (Mac) is waitlisted.
Open OpenRouter in the directory
Something wrong here? Send a correction — quote this product id: openrouter-ori-eval.