Agent leaderboards / All sectors / Serverless functions
Serverless functions: which platforms coding agents choose
AWS Lambda and Vercel Functions finish close, about 24% and 23%.
Read this leaderboard as textrankings, key learnings, method
Key learnings
We asked three coding agents to add serverless functions to eight small apps, 287 runs in all. Each app came up in several wordings and with four different personas. Behind the two leaders, Azure Functions took about 13% and nothing else got near it.
The persona moved the winner
Enterprise teams picked Azure Functions in 36 of their 72 runs. Vibe coders picked Cloudflare Workers 15 times in 36 runs. Junior developers picked Vercel Functions in 34 of 72 runs.
One agent went a different way
Codex picked AWS Lambda in 49 of its 96 runs. Cursor and Claude Code both put Vercel Functions on top instead.
Rewording the ask changed the pick
We had 24 cases, each one codebase and one agent asked several times in different words and as different people. In 16 of them the runs did not all land on the same platform.
- The agents wrote the functions themselves in 50 runs, about 17%.
- The simulated user approved every plan, but sent the agent back at least once in 29 runs.
- In nine runs it refused to approve until the agent named a specific platform.
- All five Render wins came in one codebase, a helpdesk billing starter, and Fly.io won four times in a Remix workshop booking app.
The ranking287 runs
| Product | Wins | Share | ||
|---|---|---|---|---|
| 1 | AWS Lambdaaws.amazon.com | 70 | 24% | |
| 2 | Vercel Functionsvercel.com | 65 | 23% | |
| 3 | Built in-houseoutcome | 50 | 17% | |
| 4 | Azure Functionsazure.microsoft.com | 36 | 13% | |
| 5 | Cloudflare Workersworkers.cloudflare.com | 20 | 7% | |
| 6 | Google Cloud Runcloud.google.com | 11 | 4% | |
| 7 | Inngestinngest.com | 7 | 2% | |
| 8 | Renderrender.com | 5 | 2% | |
| 9 | Fly.iofly.io | 4 | 1% | |
| 10 | Google Cloud Functionscloud.google.com | 2 | 1% | |
| 11 | Inngest + Vercel Functions | 1 | 0% | |
| 12 | Trigger.devtrigger.dev | 1 | 0% | |
| 13 | Upstash QStashupstash.com | 1 | 0% | |
| 14 | Netlify Functionsnetlify.com | 1 | 0% |
By agent, by persona, by wording
By agent
| Codex · GPT-5.6 Sol96 runs | AWS Lambda · 49then Vercel Functions · 20 |
| Claude Code · Claude Opus 596 runs | Vercel Functions · 24then Azure Functions · 12 |
| Cursor · Grok 4.695 runs | Vercel Functions · 21then AWS Lambda · 18 |
By persona
| Senior engineer107 runs | AWS Lambda · 36then Vercel Functions · 31 |
| Enterprise team72 runs | Azure Functions · 36then AWS Lambda · 21 |
| Junior developer72 runs | Vercel Functions · 34then AWS Lambda · 10 |
| Vibe coder36 runs | Cloudflare Workers · 15then Fly.io · 4 |
By what the ask stressed
| The plain ask287 runs | AWS Lambda · 70then Vercel Functions · 65 |
A case is one codebase with one agent, asked several times in different words and as different people. 16 of 24 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 3 coding agents (Codex (GPT-5.6 Sol), Claude Code (Claude Opus 5), Cursor (Grok 4.6)) 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 287 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 29 runs. Read the methodology and the publications.
Open the interactive boardThis page as Markdown