Agent leaderboards / All sectors / Deploy

Deploy: which platforms coding agents choose

Vercel took about 41% of deploy picks, Render about 34%.

270 runs6 apps3 agents2 personasupdated 2026-09-02

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

Read this leaderboard as textrankings, key learnings, method

Key learnings

We asked three agents to deploy six small apps, 270 runs in all, in several wordings and as two kinds of user. Cloudflare came third at about 11%. No other platform passed 6%.

39 vs 29

The agents did not agree on first place

We expected the agents to differ, and they did. Codex put Render first with 39 wins, ahead of Vercel at 29. Cursor and Claude Code both led with Vercel.

18 of 18

The wording changed the answer every time

A case is one codebase with one agent, asked the same thing in different words and by different people. All 18 cases flipped. None held to a single platform.

30 of 54

Cost at scale turned the order around

When we raised volume and cost, Render led with 30 of 54 runs. In the plain ask, and when we pressed on operations or on shipping fast, Vercel led.

6 of 152

Two platforms came up often and won little

Netlify was named in 152 runs and chosen in six. Fly.io was named in 144 and chosen in two.

GitHub Pages won 15 times, all on the one site that was fully static.
Google Cloud and Microsoft Azure took one win each, both on the one Python service.
The simulated user approved every plan, but sent the agent back at least once in 63 runs.
AWS Lambda came up 30 times and never won; it hosts functions, not whole apps.

Explore every run in the interactive board

The ranking270 runs

	Product	Wins	Share
1	Vercelvercel.com	112	41%
2	Renderrender.com	93	34%
3	Cloudflarecloudflare.com	29	11%
4	GitHub Pagespages.github.com	15	6%
5	Railwayrailway.app	11	4%
6	Netlifynetlify.com	6	2%
7	Fly.iofly.io	2	1%
8	Google Cloudcloud.google.com	1	0%
9	Microsoft Azureazure.microsoft.com	1	0%

By agent, by persona, by wording

By agent

Cursor · Grok 4.690 runs	Vercel · 43then Render · 27
Claude Code · Claude Opus 590 runs	Vercel · 40then Render · 27
Codex · GPT-5.6 Sol90 runs	Render · 39then Vercel · 29

By persona

Vibe coder135 runs	Vercel · 54then Render · 42
Senior engineer135 runs	Vercel · 58then Render · 51

By what the ask stressed

The plain ask108 runs	Vercel · 58then Render · 31
No operations burden54 runs	Vercel · 26then Render · 13
Volume and cost at scale54 runs	Render · 30then Cloudflare · 11
Speed to ship54 runs	Vercel · 21then Render · 19

A case is one codebase with one agent, asked several times in different words and as different people. 18 of 18 cases did not hold to a single platform.

How this was measured

Every number on this page comes from a controlled experiment. We took 6 small applications, asked 3 coding agents (Cursor (Grok 4.6), Claude Code (Claude Opus 5), Codex (GPT-5.6 Sol)) to deploy each of them, in several wordings and as a vibe coder and senior engineer, 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 270 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 63 runs. Read the methodology and the publications.

Open the interactive board This page as Markdown