AI Model Selector Helper
Get a vendor-neutral class of AI model from hard capability filters and an explained rubric — no model rankings or prices. Add your own candidates.
What do you need?
Rule-based and vendor-neutral. It recommends a class of model with reasons — no model names, rankings or prices are built in.
Choose the main task. No recommendation is made without one.
How to use this calculator
- Choose the main task, then the difficulty, response time, budget priority and any privacy or offline requirement, plus tools the model must use.
- Optionally enter or paste the typical input size (counted locally as an approximate token figure), and add the models you are considering with the facts from their documentation.
- Read the best-fit class, every rubric point behind it, the capability matrix, conflicts and trade-offs, and how each of your candidates fits.
How it works
Hard filters come first: the output the task needs (text or images), whether images must be accepted as input, and whether the model must run offline or keep data on the device. A class that fails a hard filter is excluded with its reason.
The remaining classes are ranked by a published rubric — for example +2 for the most capable tier on complex tasks, +2 for the lighter tier on a low budget, +2 for a code environment on complex coding. The score is the sum of the listed points: it measures fit to your stated needs, not model quality, and no benchmark is used.
There is no built-in list of models, prices or rankings, because they change constantly. Candidates you add are judged only on the facts you enter; anything you leave blank stays unconfirmed.
Token estimates use characters ÷ 4 and are approximate. Local does not automatically mean secure: the device, storage, licence and configuration still matter.
Worked example
Generate images with no internet. Task: generate images; privacy: must work with no internet. Only “Local image-generation tooling” passes the filters — hosted classes need a connection and text classes do not produce images. A candidate entered as a hosted text API is marked “does not fit” (produces text, not images); a workstation image tool entered as local and offline-capable fits.
Frequently asked questions
Does this use AI or send my prompt anywhere?
No. It runs entirely in your browser using rule-based checks — your prompt is never uploaded or sent to any AI model. Still, avoid pasting passwords, API keys or private data into any tool.
Why recommend a model class?
Classes such as “hosted vision-language model” or “local image-generation tooling” describe what a model must be able to do, and they do not go out of date. Pick a current product in that class and test it on your own examples.
Is the token count exact?
No. It is characters ÷ 4, a rough rule for English text. Real tokenisers differ by model and language, so treat it as approximate and check the provider's own counter for limits.
Does local mean private?
Local keeps data off a provider's servers, but privacy still depends on the device, who can access it, where files are stored and how the software is configured.
Related tools
- AI Token & Cost Estimator
- AI Context Window & Chunk Planner
- AI Privacy Redaction Helper
- AI Workflow Risk Checker
- AI Lesson Plan Prompt Builder
- Prompt Score Calculator
Explore more in AI Tools.
Cite this tool
Zerdly. (2026). AI Model Selector Helper. Retrieved from https://www.zerdly.com/tools/ai-model-selector-helper/
Tip: Enter any known values to calculate the remaining results.
Designed to run locally in your browser. Tool inputs are not intentionally stored by Zerdly unless stated otherwise.



