To compare AI models side by side, select up to three models, send your question once, and read all the answers in one place. NanthAI Edge puts the replies in a single card, stacked one under another, with each model's name next to its reply. You can then keep the answer you prefer and carry on with just that model.
How do you set up the comparison?
Tap the + button in the chat input bar and choose Participants. Search for the models you want, or use the filters for free, vision, image generation and tools. Tap each model to add it. You can select one model for a normal chat or up to three for a comparison. The selection is saved for that chat.
If you are not sure which models to pick, the "Help me choose" wizard suggests three based on your task and what matters most to you. See Participant Options for the full list of filters and sort options.
How do you send the same prompt?
Type your question once. Edge sends it to every selected model at the same time and streams each reply as it arrives. You do not have to wait for one model to finish before the next one starts.
You can also type your next message while the replies are still coming in. Edge queues it and sends it when all the replies are complete, so you do not lose your place.
What should you compare in the answers?
The replies appear one under another in the same card. Each has its own avatar, its model name and its action buttons. Read them in the same order each time, so you compare like with like.
Some things to look for:
- Accuracy. Do the facts match? If two models disagree, check the point yourself before you rely on either one.
- Instructions. Did the model do exactly what you asked, in the length and format you wanted?
- Reasoning. On a maths or planning question, does the working make sense step by step?
- Tone. For writing, which one sounds most like you or like your audience?
If you turn on web search, each model gets the search results and cites its own sources. Compare the sources too. A reply that cites a page you trust is often a better guide than one that does not.
How do you continue with one answer?
When you find an answer you prefer, fork it. That creates a branch with just that model, so you can keep going from there. The other branches stay available, and you can switch between them with the branch pill in the chat.
If one reply was weak but the others were fine, use retry-different. It swaps that one model for a new one and leaves the other replies alone. This is useful when you want a fourth opinion without starting again.
For a wider view of how the branches fit together, Ideascapes shows the whole tree as a map. Ideascapes is a Pro feature. See Branching for more.
How do you compare the costs?
Each model's reply shows its own cost once you have turned on Show Advanced Stats in Settings. A group total appears at the bottom of the card. The total is the sum of the individual replies. Different models have different rates and may produce different amounts of text, so there is no fixed multiplier for a comparison.
Costs come straight from OpenRouter, with no markup from NanthAI, according to the Edge site. See Price Transparency for how the figures are shown.
When is a comparison useful?
- Check a fact. Ask the same factual question to three models and see whether they agree.
- Draft a message. Ask for an email in three styles and pick the closest.
- Review a plan. Ask three models to find the weak points in your plan, and keep the points that more than one of them raises.
- Test a prompt. Write a prompt once and see how each model handles it before you use it in your work.
Use the same wording for every model, or the comparison tells you about your wording, not the models.
Try it
Open https://chat.nanthai.tech, select two models, and ask them one question you already know the answer to. You will see quickly how much the answers differ. For the multi-model view in more detail, see Multi-Model Chat.
