You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
If your favourite framework does not quite meet our meta-frameworks description to go in our Meta-Frameworks to add issue, you can add it to this list.
We will be expanding beyond meta-frameworks to include other setups. Please add your framework, including setup details and links, as a comment.
One thing, $mol virtualizes rendering by default, so rows outside the viewport are never created. A 1,000-row table will be measured differently than in frameworks that render all rows, so it probably needs a separate group or at least a note in the results. $mol - It for best user expirience not abt "super fast render 10 000 rows which you don't need"
I'm happy to write and maintain the mol port myself if there's interest.
Like is it approve for pull request ?
Reacted by Ivan Nesterov, Sergey Shmakov and Valik Ulyanov
Officially supports React, Preact, Vue, Solid, and Svelte. React first; the others as separate entries later, following the TanStack Start precedent
Project generation: bun create nifra my-app --template site --framework react --target node
Official scaffold as the starting point, version pinned exactly (currently 5.14.0)
Node target because the benchmark server harness runs production apps under Node. The build is still Nifra's Bun build. The scaffold's default Bun target works too if maintainers prefer it
Production server: public @nifrajs/node direct app path. The request-throughput benchmark calls the built Web app.fetch handler directly, matching the existing harness contract
Planned packages: packages/starter-nifra-react and packages/app-nifra-react, with the same measurements, routes and data shapes as the other entries
Proposal: Strapi (headless CMS), dev-time benchmarks only
Summary
Strapi is an open-source headless CMS written in TypeScript: a Node.js server with REST/GraphQL APIs and a React admin panel. It is not a rendering framework, so visitors never see Strapi-rendered pages, and it doesn't fit the meta-framework criteria in #827.
I'm proposing it for dev-time benchmarks only. Strapi is one of the most-installed Node.js backends and is very often paired with frameworks already tracked here (Next.js, Nuxt, Astro), so its install footprint and build cost are a real part of many projects' developer experience.
Build:pnpm build (strapi build: compiles TS + builds the admin panel), output in dist/
first-party-dependencies.json: the scaffold's direct @strapi/* deps: @strapi/strapi, @strapi/database, @strapi/plugin-users-permissions, @strapi/plugin-cloud
No app-strapi package: there is no public frontend to run runtime tests against
Proposed measurements
Benchmark
Applies?
Install time
✅
node_modules size / dependency count
✅
Build time
✅
Build output size
✅
First-party vs third-party dependency breakdown
✅
Runtime (Core Web Vitals, bundle size, hydration)
❌ not applicable
Preliminary local numbers (one run, pnpm 10, Node 22, not CI-comparable): ~19 s install, 784 MB node_modules, ~1,270 packages, ~30 s build (26 s of it is the admin panel), 13 MB output.
Fairness & methodology
Start from the official scaffold with default options; no tuning or trimming
Exact version pinning, same as other starters
Any benchmark-specific change gets documented in methodology.md
The methodology page will clearly mark Strapi as dev-time only / backend, so it isn't read as a frontend framework comparison
Maintenance commitment
If accepted, I'll maintain the integration: version bumps, lockfile updates, fixing build/benchmark failures, and keeping the methodology docs current.
Approval requested
Is a dev-time-only entry acceptable, given it can't run the runtime benchmarks?
If yes, should it sit with the other starters, or would you prefer a separate "backend / CMS" category?
Is the default scaffold (including @strapi/plugin-cloud) the right baseline, or should cloud be excluded?
Activity
NullVoxPopuli commentedon Nov 14, 2026
A framework I'd like to see benched:
emberjs
pnpm dlx ember.nvpDev-cmyser commentedon Nov 16, 2026
Moved to #827: $mol now has a prototype with request-time SSR on
$mol_rest_serverand client-side hydration, so it fits the meta-framework scope.A framework I'd very like to see benched:
$mol:
npx mam(dev server),npx mam path/to/app(production bundle)npm create view-tree-lsp@latest my/app(scaffolds a ready-to-run app)One thing, $mol virtualizes rendering by default, so rows outside the viewport are never created. A 1,000-row table will be measured differently than in frameworks that render all rows, so it probably needs a separate group or at least a note in the results. $mol - It for best user expirience not abt "super fast render 10 000 rows which you don't need"
I'm happy to write and maintain the
molport myself if there's interest.Like is it approve for pull request ?
TechDev009 commentedon Nov 16, 2026
A framework I'd like to see benched:
Nifra (React variant)
bun create nifra my-app --template site --framework react --target node5.14.0)@nifrajs/nodedirect app path. The request-throughput benchmark calls the built Webapp.fetchhandler directly, matching the existing harness contractpackages/starter-nifra-reactandpackages/app-nifra-react, with the same measurements, routes and data shapes as the other entriesFeedback wanted: is React an acceptable first variant, and is the official scaffold with the Node target an acceptable baseline?
unrevised6419 commentedon Nov 17, 2026
Proposal: Strapi (headless CMS), dev-time benchmarks only
Summary
Strapi is an open-source headless CMS written in TypeScript: a Node.js server with REST/GraphQL APIs and a React admin panel. It is not a rendering framework, so visitors never see Strapi-rendered pages, and it doesn't fit the meta-framework criteria in #827.
I'm proposing it for dev-time benchmarks only. Strapi is one of the most-installed Node.js backends and is very often paired with frameworks already tracked here (Next.js, Nuxt, Astro), so its install footprint and build cost are a real part of many projects' developer experience.
Context
I've been contributing e18e-style cleanups to Strapi for a while: 60+ merged PRs, including removing unused dependencies from
@strapi/strapi(strapi/strapi#27540), droppingpunycode(strapi/strapi#26189) andnodemon(strapi/strapi#27479), replacingsanitize-htmlwith DOMPurify (strapi/strapi#26150), replacing lodash with native code (strapi/strapi#27457, strapi/strapi#27454, strapi/strapi#27409), and no longer loading the admin build toolchain on every CLI command (strapi/strapi#27910). Tracking Strapi here would give a neutral, ongoing measurement of its footprint over time.References
Proposed setup
packages/starter-strapi(ownpackage.json+ lockfile, not in the pnpm workspace)pnpm-workspace.yaml@strapi/strapipinned to5.62.0(exact)pnpm build(strapi build: compiles TS + builds the admin panel), output indist/first-party-dependencies.json: the scaffold's direct@strapi/*deps:@strapi/strapi,@strapi/database,@strapi/plugin-users-permissions,@strapi/plugin-cloudapp-strapipackage: there is no public frontend to run runtime tests againstProposed measurements
node_modulessize / dependency countPreliminary local numbers (one run, pnpm 10, Node 22, not CI-comparable): ~19 s install, 784 MB
node_modules, ~1,270 packages, ~30 s build (26 s of it is the admin panel), 13 MB output.Fairness & methodology
methodology.mdMaintenance commitment
If accepted, I'll maintain the integration: version bumps, lockfile updates, fixing build/benchmark failures, and keeping the methodology docs current.
Approval requested
@strapi/plugin-cloud) the right baseline, or should cloud be excluded?I won't open a PR until this is approved.