Last benchmarked October 20
Sandbox hardware and software context — use this to tell whether a provider is faster because of more CPUs/memory, not a faster disk or network.
Median workload time x published rates at the 8 vCPU / 16 GiB sandbox spec. Excludes sandbox create/destroy time.
$0.04032/vCPU-hr + $0.0129/GiB-hr. Custom sizing needs Hobby plan or above. Source · Compare all providers
Total disk usage (du -sx) captured after each phase — a proxy for disk I/O and storage cost. Some providers report 0 here if their sandbox filesystem doesn't support usable disk-usage reporting.
Each iteration runs the full clone + install + typecheck cycle from a fresh sandbox.
| # | Phases | Total | Prepare | Bun DL | Bun Unpack | Clone | Install | Typecheck | Error |
|---|---|---|---|---|---|---|---|---|---|
| 1 | 6/7 | 302.08s | 18.56s | 1.26s | 1.53s | 5.67s | 34.71s | — | script failed after install (exit code 124, 6/7 phases completed): Saved lockfile | $ bun run --cwd packages/core fix-node-pty | $ bun run script/fix-node-pty.ts | $ husky | $ bun turbo typecheck | @opencode-ai/effect-drizzle-sqlite:typecheck: $ tsgo --noEmit | @opencode-ai/httpapi-codegen:typecheck: cache miss, executing 81c22208e92d78a8 | @opencode-ai/httpapi-codegen:typecheck: $ tsgo --noEmit | @opencode-ai/tui:typecheck: cache miss, executing f8bcdb739b90c436 | @opencode-ai/tui:typecheck: $ tsgo --noEmit |