Choosing which model builds which part of a prototype is becoming a design decision. The designers who test, compare and record that choice will move fastest.
Gemini 4 Argon in Figma Make: the short answer
Gemini 4 Argon in Figma Make is worth trying when your prototype needs depth, motion or 3D that the default model handles poorly, and not worth switching to for every prompt. Figma added Google's newest model to the Make model picker on 13 October 2026, on all plans, according to the Figma release notes. You pick it from the model name in the Make chat box, and you can switch back at any point in the same conversation.
The useful question is not "is Argon better" but "which parts of my prototype should a different model build". This playbook gives you a simple loop for that, called Pick, Probe, Pin, plus the settings, prompts and checks that matter before anything you make with it reaches a stakeholder or a developer.
What Figma and Google actually shipped
Kept to what the official pages state.
Where it lives. Figma's help article Select an AI model to use in Figma Make explains the steps: open a Make file, click the model name shown in the chat box, and choose a model from the dropdown. Your team or organization must have its AI setting turned on.
What else is in the picker. The same article lists Gemini 4 Argon next to several OpenAI, Anthropic and Google models, plus a default model that every file starts with. The menus, chat panel and code view stay the same whichever model you choose.
How files remember. New files start on the default model. An existing file reopens on the model you last selected for it. That detail matters for teams, because a shared file can quietly be running on a model a teammate picked.
Credits. Figma says model choice affects AI credit use, that non-default models can use substantially more credits per task, and that pricing for them is still being worked out. Argon is on a promotion that uses fewer credits until 30 November 2026, after which usage rises. Check your own credit history rather than trusting any estimate, including the one in this article.
Known rough edges. Figma warns that non-default models may show "unexpected errors, additional self-healing, or longer reasoning and generation." Treat that as a real caveat, not legal boilerplate.
What the model is for. Google's Gemini 4 Argon announcement focuses on visual coding, 3D scenes, interactive animation and stronger visual understanding, and says the model follows style instructions more closely than earlier Gemini models. The Gemini 4 Argon guide for designers published at launch noted that the model shipped with no Figma integration. That gap closed the same day, through Make.
Hype vs reality
The hype says you can now prompt a 3D product site inside Figma and hand it to engineering. That overstates it.
What holds up: Make already turns prompts into working code, and a model that Google positions for 3D, shaders and interaction gives you a second option when the default model produces flat or stiff results. Being able to switch per prompt, inside one file, makes comparing models cheap.
What does not hold up: a prototype is still a prototype. A strong WebGL hero scene says nothing about performance on a mid-range phone, accessibility for people who turn off motion, or whether your engineering team wants Three.js in production. Figma itself warns about errors and longer generation on non-default models. The verdict: use Argon as a specialist for motion and spatial moments, keep the default model for layout and flows, and judge both on the same brief.
Default model vs Gemini 4 Argon in Make
Situation | Stay on the default model | Switch to Gemini 4 Argon |
|---|---|---|
Standard screens, forms, dashboards | Yes, predictable and lower credit use | Rarely worth it |
3D hero, product viewer, scroll scenes | Often flat or simplified | Try it first, this is its stated strength |
Micro-interactions and animated states | Fine for simple transitions | Useful for physics, easing and layered motion |
Matching a strict visual style | Depends on your prompt and references | Google says style following improved, so test it |
Credit budget is tight | Safer choice | Cheaper during the promotion, check after 30 November |
Shared team files | Clear default for everyone | Name the model in the file so nobody is surprised |
The Pick, Probe, Pin loop
A simple way to bring a new model into Make without wasting credits or confusing your team.
Pick: decide which part of the prototype needs it
Do not start a whole file on Argon. Build the structure, navigation and content on the default model first. Then mark the moments that carry the experience: a 3D product turntable, an animated onboarding sequence, a data visual that responds to input. Those are the candidates.
Probe: run the same brief on both models
Write one prompt and send it twice, once per model, in separate files or after duplicating the file. A brief that works well for spatial work reads like direction for a motion designer:
"Build a hero section with a 3D rotating headphone model on a soft studio backdrop. Slow orbit, pause on hover, drag to rotate. Respect reduced motion by showing a static render."
"Keep the existing type scale and colors from this file. Do not change navigation or layout outside the hero."
"Explain in the chat which libraries you used and anything that may run slowly on mobile."
Compare what you get on four things: does it match the brief, does it respect the constraints, does it run smoothly in preview, and how many credits did it cost. Write the result down in the file description. A two-line note saves your team from repeating the experiment.
Pin: lock the decision and label it
If Argon wins for a section, keep using it for that kind of task and label the file or page ("Hero built with Gemini 4 Argon, rest on default"). Because a file remembers the last model picked, anyone opening it next will be on Argon unless they change it. If the default model wins, switch back and move on. Repeat the probe when Figma changes the default model, which its help article says can happen over time.
What to check before you share the output
Open the code view and look at what was added. Unfamiliar 3D or animation libraries are a conversation to have with engineering before a stakeholder falls in love with the demo. Preview on a real phone, not only the desktop canvas. Check reduced motion, keyboard focus and contrast on anything animated. And check the credit history after a heavy session, since a few long generations can cost more than you expect.
If the prototype will become a real product, treat it as a spec with a live demo attached. The guide to Figma code layers for designers covers how working prototypes move closer to production, and AI prototyping explains where prompt-to-prototype fits in discovery.
Where design work is going
Model pickers inside design tools are becoming normal. Over the next one to three years, choosing which model builds which part of a prototype will feel like choosing a font or a component library: a craft decision with trade-offs, not an engineering detail. Designers who can say "this scene needs a model that is strong at 3D, this flow needs a predictable one" will move faster than those who accept whatever is selected.
The bigger shift is that motion and spatial work stop being expensive extras. When a convincing interactive hero is one prompt and a review away, more teams will try it, and the job moves toward direction, restraint and performance budgets. The piece on motion design for AI goes deeper on that.
The fear: am I falling behind if I do not know which model to use?
No. Nobody has a stable answer, because the list keeps changing and Figma itself says credit pricing is still in development. What matters is having a method. If you can run the same brief on two models, compare results against clear criteria and record the decision, you are ahead of most teams, whatever names are in the dropdown next quarter.
The honest risk sits elsewhere. Designers who only produce static mockups of motion and 3D will feel pressure, because a working version is now cheap to generate. Concrete moves: build one interactive moment in Make this month, learn enough about Three.js and animation libraries to read what the model wrote, and make reduced motion and mobile performance part of every review. The answer to will AI replace UX designers explains why judgment is the part that stays.
Getting help
If you want spatial or motion-heavy prototypes turned into products that hold up in production, see DIGCY motion design and AI video and product design services. DIGCY is a Dribbble Selected Agency and a UI/UX and product design studio in Morocco, serving Casablanca and working with clients globally.
FAQ
How do I use Gemini 4 Argon in Figma Make?
Open a Figma Make file, click the model name in the chat box, and choose Gemini 4 Argon from the dropdown. Figma says it is available on all plans as long as your team or organization has AI turned on. You can switch back to the default model at any point in the conversation.
Does Gemini 4 Argon cost more AI credits in Figma Make?
Figma says non-default models can use more credits and that pricing for them is still in development. Gemini 4 Argon is on a promotion with lower credit use until 30 November 2026. Check your credit history after a few prompts to see the real cost for your work.
When should designers pick Gemini 4 Argon over the default model?
Pick it for 3D scenes, product viewers, layered animation and interactive visuals, which match what Google says the model is built for. Keep the default model for standard screens and flows, where predictability and lower credit use matter more.
Is Gemini 4 Argon in Figma Make overhyped?
Partly. Having a model aimed at 3D and motion inside Make is useful, and switching per prompt makes comparison easy. But Figma warns of errors and slower generation on non-default models, and a prototype still needs performance, accessibility and engineering review before it becomes a product.
Will AI models in Figma Make replace UI designers?
No. The model writes code from a brief. Someone still decides what the experience should be, which model to use for which part, and whether the result is good enough to ship. Those decisions are design work, and they become more important as generation gets cheaper.
Let’s keep in touch.
Discover more about high-performance web design. Follow us on Twitter and Instagram.
