Help & FAQ
Everything you need to get the most out of Prism
What is Prism?
Prism runs your prompt against multiple AI models at once and shows their answers side by side, so you can compare quality, style, and reasoning without copy-pasting between tabs. Pick the models you want in the Models section of the sidebar, type a question, and run.
What is a run?
A run is one prompt sent to your selected models. In Compare and Debate a single run fans out to every selected model in parallel; in Single it goes to one. Each run adds a turn to the conversation thread. Follow-ups and re-runs each count as their own run.
Compare vs. Debate vs. Single
Single sends your prompt to one model. Compare sends it to every selected model and lays the answers out in a grid. Debate goes further: after the models answer, Prism analyzes their disagreements, synthesizes the strongest view, and fact-checks claims. The mode switches automatically as you add or remove models, and you can change it from the composer.
Smart Routing
When you're in Single Model mode, Smart Routing analyzes your prompt and recommends the cheapest model that can handle it well — suggesting an upgrade only when the task needs a stronger model. Toggle it from the sidebar footer. It's available in Single mode only, since Compare and Debate are about running several models on purpose.
Continuing a conversation
Each run adds a turn to the thread; new answers appear below the previous ones — scroll up to revisit earlier turns. Models you add mid-conversation only run on the new prompt, not retroactively. To dig into one model's answer, use the Ask box under it to open a focused follow-up thread with just that model — it carries the original question and that answer as context.
Workspaces & grounding
Turn on Workspace Mode and upload documents to ground answers in your own material. Mark a document Confidential to keep it out of any training data while still using it for grounding.
Web grounding
For time-sensitive questions — news, scores, prices, “latest”, a recent year — Prism automatically searches the web and feeds fresh results to the models so they answer from current information instead of stale training data. There's no toggle: it fires only when your question looks time-sensitive. When it does, a Grounded in web results chip appears above the answers with the source count, and the sources are listed below them. Everyday questions skip the search and stay fast.
Incognito
Toggle Incognito (the eye icon in the sidebar header) to run privately: those comparisons aren't saved to your history and aren't used for training. Runs from an incognito session are discarded when you close the tab unless you save them.
Which models are included
Models are grouped into tiers in the sidebar. Free covers fast, low-cost everyday models. Pro adds stronger mid-tier models, and Power unlocks the frontier models. Max includes everything plus the largest monthly credit allowance. The Models section always shows the full roster; anything above your plan appears with a lock and an Upgrade button. You can select any mix of the models your plan allows.
Plans, credits & billing
Your plan — Free, Pro, Power, or Max — shows on your profile badge in the sidebar and determines which models you can run. Upgrading unlocks the higher tiers immediately; you can change or cancel anytime from Settings, which opens the billing portal to manage your subscription and view invoices.
Troubleshooting
If one model shows an error while the others answer, it's usually a temporary provider hiccup or rate limit — re-run the prompt or retry that turn. If every model fails identically, the service is likely having a brief outage; try again shortly. Answers that seem out of date? Web grounding only fires for time-sensitive questions — add a cue like “latest” or a year to trigger it. Can't select a model? It's above your current plan — upgrade to unlock it.
Contact support
Email us at support@prism-engine.ai and we'll get back to you. You can also review our Privacy Policy and Terms of Service.