Document large GPU browsers

[yummybomb]

Summary

Documents the large GPU browser tier, requested by setting memory to 12GiB together with gpu: true.

• browsers/gpu-acceleration.mdx gets a new "Large GPU browsers" section covering:
  • the standard vs. large allocation (4 vCPU / 6 GiB / 2 GiB VRAM vs. 8 vCPU / 12 GiB / 4 GiB VRAM)
  • TypeScript, Python, and Go examples
  • the rule that 12GiB is the only memory value accepted with gpu: true
  • the HTTP 529 no_gpu_browsers_available response when no large GPU browser is available
• browsers/process-execution.mdx: the whole-session memory paragraph now covers GPU allocations.

Notes

• Merge after the API change that accepts memory: 12GiB for GPU browsers is deployed.
• The Go example passes the raw string "12GiB", like the pools page does. It compiles against the current Go SDK (v0.120.0), where BrowserMemoryRequest is a string type. The TypeScript and Python SDKs type memory as a literal union, so those examples only type-check once the SDKs are regenerated with the new enum value. At runtime they send the string either way.
• Pricing for the large tier is not stated here. Add a line to this page and to info/pricing.mdx once the rate is confirmed.

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated
Kernel	🟢 Ready	View Preview	Oct 6, 2026, 10:01 PM

💡 Tip: Enable Automations to automatically generate PRs for you.