# Model Choice

> Pick a model for an agent by quality, speed, cost and where data may go - and say why.

- Skill: `vstorm-co/model-choice` (Agent Skill)
- Install (CLI): `npx get-skill add vstorm-co/model-choice`
- Raw SKILL.md: https://api.skillmd.com/api/skills/vstorm-co/model-choice/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: vstorm-co (https://skillmd.com/u/vstorm-co)
- Updated: 2026-11-12
- Page: https://skillmd.com/skills/vstorm-co/model-choice

---


# Choosing a model

Look up current prices and models with web search before recommending one -
they change monthly. Cite the page.

## The four questions

1. **How hard is the job?** Reasoning over several documents, writing code or
   planning steps needs a frontier model. Classifying, extracting or short FAQ
   answers run well on a small, cheap one.
2. **How fast must it answer?** A chat widget wants a quick model; a nightly
   report does not care.
3. **How much will it run?** Price × messages per month. Estimate it out loud.
4. **Where may the data go?** Some companies need a model in their own region,
   their own cloud, or open weights they run themselves.

## Recommend

One model, the reason in a sentence, and the cheaper fallback: "Claude Sonnet for
the answers; Haiku would be a third of the cost if quality holds." Offer to
switch later - the model is one setting.

## Never

Recommend a model the organization has no key for without saying how to add one.

