Agent leaderboards / All sectors / Serverless functions

Serverless functions: which platforms coding agents choose

AWS Lambda and Vercel Functions land almost even, about 24% and 23%.

476 runs8 apps5 agents4 personasupdated 2026-09-29

The interactive board, open on serverless functions. Open it full page · this page as Markdown

Read this leaderboard as textrankings, key learnings, method

Key learnings

We asked five coding agents to add serverless functions to 8 codebases, 476 runs in all. Each codebase got the same ask in several wordings, and from four kinds of user. Third place went to Azure Functions, at about 12%, and no other platform got near it.

59 of 120

Who asks changes which platform wins

Enterprise teams picked Azure Functions in 59 of their 120 runs. Junior developers picked Vercel Functions 57 times. Vibe coders picked Cloudflare Workers 23 times.

26 of 40

The same ask, worded again, got another answer

A case here is one codebase with one agent, asked several times over. In 26 of 40 cases the runs did not all settle on the same platform.

1 of 118

Named in many runs, chosen once

Netlify Functions came up in 118 runs and was picked in one of them. Upstash QStash came up in 69 runs and was picked twice.

  • Codex picked AWS Lambda in 67 of its 144 runs.
  • Cursor, Claude Code and Muse Code each picked Vercel Functions most often.
  • In 88 runs the agents named no platform and built it in-house, about 18%.
  • The simulated user approved every plan, but sent the agent back at least once in 66 runs, and in 17 refused until a product was named.
  • All eight Fly.io wins came in one codebase, a Remix workshop booking platform.
Explore every run in the interactive board

The ranking476 runs

ProductWinsShare
1 AWS Lambdaaws.amazon.com 114 24%
2 Vercel Functionsvercel.com 110 23%
3 Built in-houseoutcome 88 18%
4 Azure Functionsazure.microsoft.com 59 12%
5 Cloudflare Workersworkers.cloudflare.com 30 6%
6 Google Cloud Runcloud.google.com 14 3%
7 Inngestinngest.com 8 2%
8 Fly.iofly.io 8 2%
9 Trigger.devtrigger.dev 8 2%
10 Renderrender.com 7 1%
11 Google Cloud Functionscloud.google.com 4 1%
12 Upstash QStashupstash.com 2 0%
13 Inngest + Vercel Functions 1 0%
14 Netlify Functionsnetlify.com 1 0%
15 AWS Fargateaws.amazon.com 1 0%

By agent, by persona, by wording

By agent

Codex144 runsAWS Lambda · 67then Vercel Functions · 32
Claude Code120 runsVercel Functions · 30then Azure Functions · 15
Cursor119 runsVercel Functions · 26then AWS Lambda · 22
Muse Code · Muse Spark 1.371 runsVercel Functions · 18then AWS Lambda · 10
Grok Build CLI · Grok 4.722 runsAWS Lambda · 6then Vercel Functions · 4

By persona

Senior engineer177 runsAWS Lambda · 57then Vercel Functions · 53
Enterprise team120 runsAzure Functions · 59then AWS Lambda · 40
Junior developer119 runsVercel Functions · 57then AWS Lambda · 14
Vibe coder60 runsCloudflare Workers · 23then Fly.io · 8

By what the ask stressed

The plain ask476 runsAWS Lambda · 114then Vercel Functions · 110

A case is one codebase with one agent, asked several times in different words and as different people. 26 of 40 cases did not hold to a single platform.

How this was measured

Every number on this page comes from a controlled experiment. We took 8 small applications, asked 5 coding agents (Codex, Claude Code, Cursor, Muse Code (Muse Spark 1.3), Grok Build CLI (Grok 4.7)) to add serverless functions to each of them, in several wordings and as a senior engineer and enterprise team and junior developer and vibe coder, and let the agent choose the product. Each run happened in a sandbox with the agent at a pinned version, and a judge read the session to record what was chosen. That is 476 runs. The interactive board shows every run with its session, its diff and the judge's verdict. A simulated user stood in for the owner of the codebase: it read the agent's plan and had to approve it before any code was written; it sent the agent back at least once in 66 runs. Read the methodology and the publications.

If you sell in this sector: what these numbers mean for a vendor.

Open the interactive boardThis page as Markdown