Can AI build 3D product visuals from code alone?
Both top models can now build a recognisable 3D object from nothing but code, which makes spinning product visuals and interactive demos cheap to try.
What the reel showed
One brief to two AI agents: build a 3D Rick Sanchez out of nothing but code. No model files, no images, no textures. The spiky hair, the unibrow, the lab coat and the flask, all from shapes each model wrote itself, turning on a turntable in one web page.
Claude Fable 5.1 in Claude Code lit a single figure on a dark stage, in about 7 minutes for $2.99. GPT-6 Astra in Codex CLI put its figure inside an 'interdimensional archive' card, in under 3 minutes for $0.62.
Both are recognisable. Neither loaded a single asset.
| Entrant | Tool | Time | Cost | Verdict |
|---|---|---|---|---|
| Claude Fable 5.1 | Claude Code | 7m09 | $2.99 | Single lit figure on a dark stage |
| GPT-6 Astra | Codex CLI | 2m54 | $0.62 | Figure in an archive card |
Each model ran in its maker's coding agent with the same brief, one attempt each, and each cost is read off that run's own bill.
What this means for your business
Code-built 3D is how you get a product visual that spins, zooms and opens up without hiring a 3D artist or loading heavy files. It runs in any browser and weighs almost nothing.
Uses: a product viewer that shows your colours and options, an exploded view of what is inside the box, an interactive diagram of how your service works, a light animated hero on a landing page, a demo to share on a sales call.
On this kind of work the gap between the top models is now mostly taste. Run the same brief through both, look at both, and pick on the result.
What I’d use
For a quick visual I would start with Astra. It was about 4.8 times cheaper here and just as recognisable. If the piece is going on a client's homepage, I run both and pick the better one. At these prices that is cheap insurance.
Give both exactly the same instructions. Otherwise you are comparing your prompts, not the models.
Day to day I run both on subscriptions, not per call: Fable in Claude Code on my Claude plan, Astra in Codex CLI on my ChatGPT plan.
Try it yourself
Get a spinning 3D version of one of your own products, built from code, and compare two models on it. Use the subscriptions you may already have, or one OpenRouter account.
Route A: Claude Code on your Claude plan
This is how I work every day. Claude Code is included with Claude Pro and Max, so runs draw on your plan's usage limits instead of a bill per call. Fable is included on Max; on Pro it runs on paid usage credits.
Route A: Codex CLI on your ChatGPT plan
Codex CLI is OpenAI's coding agent. Run it and choose Sign in with ChatGPT to use it on a ChatGPT Plus or Pro plan. GPT-6 Astra is available in Codex on both.
Route B: OpenRouter, both in one window
No subscription, or you want both answers side by side? One OpenRouter account and key covers Claude, GPT and hundreds of other models, pay per use. Open the chat and add Claude Fable 5.1 and GPT-6 Astra, so the same message goes to both.
Describe your product
Be concrete about shape, size and the details people recognise it by. On Route A, start each agent in its own empty folder and paste the same brief into both.
Build a 3D model of [your product, e.g. "our 500ml stainless steel water bottle with a bamboo lid"] as a single HTML file called product.html. - Use three.js loaded from a CDN. No model files and no images: build it from basic shapes. - Get the proportions and the details that make it recognisable right. - Slow turntable spin, soft studio lighting, plain background. - Drag to rotate, scroll to zoom. - Three buttons that change the main colour: [your colours]. If you can write files, save it in this folder. Otherwise return the whole file in one code block.
Open it in your browser
On Route A the file is already in each folder: double-click product.html. From the chat, copy each answer into a plain text file, save it as product.html and double-click it. Keep your internet on, because the page loads three.js.
Pick on the result
Which one looks like your product? Which would you put on your website? On OpenRouter, your activity page shows what each answer cost.
Something I haven't tried yet, but you can
Higgsfield, an AI image and video platform, has an official MCP connector for AI agents. In Claude Code you can ask the agent to set up Higgsfield for you, and Codex CLI can add it as an MCP server. Then the same agent that built your 3D product could generate product shots, ad variations or a short promo clip from it. It uses a paid Higgsfield plan and its credits. I have not tested this, so treat it as an idea to try.
Links
- Claude CodeAnthropic's coding agentclaude.com
- Claude Code with Pro or MaxRun it on your Claude subscriptionsupport.claude.com
- Codex CLIOpenAI's coding agent, signs in with a ChatGPT plangithub.com
- OpenRouter chatSend one prompt to several models side by sideopenrouter.ai
- Claude Fable 5.1 on OpenRouteropenrouter.ai
- GPT-6 Astra on OpenRouteropenrouter.ai
- three.jsThe free 3D library both models usedthreejs.org
- Higgsfield MCPHiggsfield's official connector for AI agentshiggsfield.ai
Want an interactive product visual on your site?
A code-built 3D view loads fast and costs a fraction of a 3D shoot. I can build it properly and fit it into your site.