---
name: minirouter-choose-model
description: Chooses compatible MiniRouter model and effort candidates using public prices and exact benchmark evidence. Use when selecting an LLM, comparing model cost or choosing a model for a task without signup.
compatibility: Node.js 22 or newer with filesystem access. Online commands require HTTPS access to minirouter.sh and api.minirouter.sh.
---

# Choose a model

**IS:** public model selection and cost estimates. **IS NOT:** measured task evaluation or changing a user's routing.

Resolve every relative path against this installed SKILL.md directory. Run the bundled `scripts/model-advisor.mjs`; do not recreate its calculations.

1. Translate the user's task into intelligence, coding or math. Capture context/output capacity, required input modalities and API parameters, expected disjoint token counts, and any provider/budget constraint. Read [the input contract](references/input.md) when writing requirements; start from `examples/requirements.json`.
2. Run `node <skill-directory>/scripts/model-advisor.mjs choose --input <requirements.json> --out <new-output-directory>`. No API key is needed. A saved evidence file can be supplied with `--catalog <snapshot.json>` for a historical offline comparison.
3. Read report.json and report.md. Return up to three exact model/effort configurations, estimated cost, evidence date and the main tradeoff. Link the saved report. If evidence is insufficient or unavailable, state the missing evidence rather than inventing a winner.

Continue until the command succeeds and the report answers the user's constraints, or identify the exact missing input/source. Reading public evidence and writing the requested local report need no additional confirmation. The skill does not send inference, create an account or apply a configuration.

## Gotchas

- A benchmark variant belongs to its exact reasoning effort. An unknown effort is not default.
- A request minimum, unknown price or missing cache semantics can make the full estimate unavailable.
- Archived evidence cannot establish today's availability. Benchmarks do not measure MiniRouter route latency or the user's task success.
- Preserve the returned provider restriction and effort in any proposed configuration.

Use minirouter-audit-spend for local usage files and minirouter-evaluate-switch for measured comparisons.

## Install this release

Download the [complete skill archive](https://minirouter.sh/agent-skills/releases/83b1177c2858898aef4af631fd5d5060d3afcac0c17b4719d19370b084d57721/minirouter-choose-model.tar.gz) and [release manifest](https://minirouter.sh/agent-skills/releases/83b1177c2858898aef4af631fd5d5060d3afcac0c17b4719d19370b084d57721/manifest.json). Verify the archive SHA-256 against the manifest before extracting into a new local directory. Keep the full extracted skill folder; SKILL.md alone does not include the executable. Run `node <skill-directory>/scripts/model-advisor.mjs --help` to verify Node 22+ installation.
