Agent leaderboards / All sectors / Deploy

Deploy: which platforms coding agents choose

Vercel and Render finished almost even, about 39% and about 38%.

508 runs6 apps5 agents2 personasupdated 2026-09-29

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 five coding agents to deploy six codebases, over 508 runs. We asked in different words and as different people. Cloudflare came third at about 12%, and the rest of the field was thin.

67 of 150

The agents didn't agree on the leader

We expected the agents to differ, and they did. Cursor and Claude Code picked Vercel most often. Codex, Muse Code and Grok Build CLI picked Render most often, though the last two were near ties. Codex had the widest gap, 67 of its 150 runs.

27 of 30

The wording changed the answer

Take one codebase and one agent, then ask over and over in different wordings. In 27 of 30 such cases the runs didn't all land on the same platform.

57 of 102

Cost at scale moved the order

Asked about volume and cost at scale, Render led with 57 of 102 runs. On the plain ask, Vercel led with 97 of 204.

9 of 249

Two platforms were named often and picked rarely

Netlify came up in 249 runs and won nine of them. Fly.io came up in 218 runs and won four.

  • The simulated user, who approves each plan before code is written, sent the agent back at least once in 162 runs.
  • GitHub Pages won 26 times, all on the one site that was fully static.
  • AWS Lambda was named in 44 runs and never picked, and it hosts functions rather than whole apps, so it wasn't really an alternative.
Explore every run in the interactive board

The ranking508 runs

ProductWinsShare
1 Vercelvercel.com 197 39%
2 Renderrender.com 193 38%
3 Cloudflarecloudflare.com 59 12%
4 GitHub Pagespages.github.com 26 5%
5 Railwayrailway.app 17 3%
6 Netlifynetlify.com 9 2%
7 Fly.iofly.io 4 1%
8 Google Cloudcloud.google.com 1 0%
9 Microsoft Azureazure.microsoft.com 1 0%
10 Tailscaletailscale.com 1 0%

By agent, by persona, by wording

By agent

Codex150 runsRender · 67then Vercel · 50
Claude Code120 runsVercel · 51then Render · 36
Cursor119 runsVercel · 52then Render · 41
Muse Code · Muse Spark 1.389 runsRender · 37then Vercel · 34
Grok Build CLI · Grok 4.730 runsRender · 12then Vercel · 10

By persona

Vibe coder254 runsVercel · 95then Render · 89
Senior engineer254 runsRender · 104then Vercel · 102

By what the ask stressed

The plain ask204 runsVercel · 97then Render · 67
Volume and cost at scale102 runsRender · 57then Cloudflare · 20
No operations burden101 runsVercel · 49then Render · 28
Speed to ship101 runsRender · 41then Vercel · 37

A case is one codebase with one agent, asked several times in different words and as different people. 27 of 30 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 5 coding agents (Codex, Claude Code, Cursor, Muse Code (Muse Spark 1.3), Grok Build CLI (Grok 4.7)) 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 508 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 162 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

Other sectors