A founder's product lets customers upload a 6 MiB image for background removal. It works locally, yet the deployed endpoint returns 413 before application code sees the file. On November 20, 2026, Vercel announced a public beta raising inbound Function request bodies to 20 MiB, up from the usual 4.5 MB. That can simplify a small-file workflow, but the opt-in has plan, runtime, and memory conditions.
Check the conditions before changing your upload path
Set VERCEL_SUPPORT_LARGE_REQUEST_BODY=1 in the environment where you want the beta. The setting applies to the whole project within that environment, so Preview is a sensible first stop. Your Functions must use the 4 GB Performance memory setting. Vercel's memory documentation lists that setting for Pro and Enterprise; Hobby remains at 2 GB. The larger body size applies to Node.js, Python, and Bun Functions and to Routing Middleware.
Edge-runtime Functions and Routing Middleware retain the 4.5 MB cap. A 6 MiB request to an Edge route can still return 413. For a Next.js route that needs the beta, Vercel says to use the Node.js runtime with runtime: 'nodejs', or review the use of proxy.ts. There is a second architectural change: once the project opts in, Routing Middleware sees headers and metadata but not the request body. Move any body-dependent checks out of middleware before enabling it.
Try a disposable 6 MiB request in Preview
For a Next.js app, make app/api/body-check/route.ts a diagnostic route that counts incoming bytes without saving them. This is a documentation-based example; we have not deployed it to a Vercel account for this article.
export const runtime = 'nodejs';
export async function POST(request: Request) {
if (!request.body) return new Response('Missing body', { status: 400 });
let bytes = 0;
for await (const chunk of request.body) bytes += chunk.byteLength;
return Response.json({ receivedBytes: bytes });
}
Set the project's Preview Functions to Performance 4 GB, add the beta environment variable for Preview, and deploy a branch. Then create a harmless test file and send it to the preview URL:
python3 -c "open('sample-6m.bin','wb').write(b'0' * (6 * 1024 * 1024))"
curl -i --data-binary @sample-6m.bin \
-H 'Content-Type: application/octet-stream' \
https://YOUR-PREVIEW.vercel.app/api/body-check
If the request reaches the Function, the response should report receivedBytes: 6291456. Compare it with an Edge route or a Preview without the opt-in to recognize the 413 path. A successful byte count proves only that this payload crossed the platform boundary. It does not validate file type, protect the route, store the image, or establish safe memory use. Remove the diagnostic route before production.
Decide whether the extra memory is worth it
Vercel says to read large bodies promptly and avoid holding them in memory. Performance memory can also change the bill: Function charges depend on active CPU, provisioned memory, region, and time. Check the current pricing documentation and your own traffic. If you expect larger videos or many uploads, the 413 guidance recommends client-side uploads directly to Vercel Blob so files do not pass through your Function. Start with a real file and a Preview deployment, then choose between the beta and direct storage based on measured cost, security checks, and user experience.
