Which AI to choose for your use? This four-question questionnaire recommends three AI models based on your task, budget, confidentiality needs, and required context size. The calculation is done by readable rules and weightings, without AI, in your browser.
How to Use It
- Choose your main task: writing, coding, summarizing long documents, data analysis, translation, agents, customer support, information retrieval, images and scanned documents, or reasoning.
- Indicate your budget: free, cost-controlled API, or no constraint.
- Specify confidentiality: general use, sensitive data, or requirement to stay local.
- Add if needed a minimum context size, image reading, or tool calls.
- Read the three recommendations and their reasons: the result updates with each change and the page link preserves your answers, so you can share it.
How the Three Models Are Chosen
Hard constraints filter models first: image reading, tool calls, context size, open weights if you must stay local, free if your budget requires it. Each remaining model then gets a score: its quality index, a bonus for reasoning and tool calls depending on the task, minus a price penalty which weighs more if your budget is controlled. The three recommendations are the best score, the cheapest model among those whose quality remains close to the best, and the best open-weights option. The weights per task are written by hand and editable in the plugin administration.
To compare more models side by side, use the model comparator; to estimate the cost of retained usage, the API cost calculator or, for an agent, the agent cost calculator.
Limitations to Know
Only models with a quality index are ranked, and they are a minority: others are never recommended here, for lack of measurement. This index is general: our data does not contain scores by domain (code, mathematics, agents), so the task acts on conditions and weightings, not on a task-specific measurement. We have no data on response speed, content language, or hosting location: ‘open weights’ does not mean ‘European hosting’. Prices are OpenRouter’s public prices of the day, with a mixed price of 3 parts input for 1 part output. If no model meets all conditions, the tool relaxes one constraint and indicates it.
Frequently Asked Questions
Are my answers sent anywhere?
No. The calculation is done in your browser. Your answers are only saved in the page address, so you can share the link.
Why isn’t my preferred model suggested?
Either it doesn’t have a quality index in our data, or it doesn’t meet one of your conditions (context, vision, tools, free, open weights). Relax a condition to see it appear.
Why isn’t the second model always ‘best price’?
When the best choice is already the cheapest among models of similar quality, the second is not cheaper: it is then presented as a similar-quality alternative, not as the best price.