TL;DR: The ranked recommendation by use case, so you can skip the next 6,000 words.

  • Content sites, marketing pages, blogs, programmatic SEO: Astro 5. Ship static HTML, hydrate only what needs JavaScript, Lighthouse scores of 95-100 out of the box. Cheapest to host by a wide margin.
  • SaaS apps, dashboards, authenticated products: Next.js 16 with the App Router. Largest hiring pool, deepest ecosystem, Server Components are finally stable.
  • Smallest possible bundle, highest performance ceiling: SvelteKit with Svelte 5 runes. Your team needs to be comfortable outside the React mainstream.
  • Vue shops, especially EU/Asia teams: Nuxt 3. It's the Vue meta-framework, period.
  • URL-centric products (marketplaces, multi-step forms): Remix (now React Router 7). Web-standard forms, progressive enhancement by default.
  • Gatsby: Migrate off it. Seriously.

What "Modern Framework" Actually Means in 2026

When a CTO or marketing leader says "framework" in 2026, they're usually conflating three or four layers into one word. Let me break that apart because the distinction matters when you're signing a contract or hiring a team.

A renderer is the UI library: React 19, Vue 3, Svelte 5, or Preact. It decides how your components turn into DOM nodes.

A meta-framework is the layer on top: Next.js, Astro, SvelteKit, Nuxt, Remix. It handles routing, data loading, server rendering, static generation, and the build pipeline. This is the layer where 90% of your architectural decisions live.

A bundler is the tool that compiles, minifies, and chunks your code. In 2026, that's Turbopack (Next.js), Vite (Astro, SvelteKit, Nuxt, Remix), or Webpack (legacy projects still hanging on).

A deploy target is where the output runs: Vercel, Netlify, Cloudflare Pages, AWS Amplify, a plain VPS, or your own Docker containers. The meta-framework dictates which deploy targets are first-class citizens.

Here's why this matters: choosing a "framework" in 2026 is not like choosing WordPress vs Squarespace in 2015. It's an architecture decision that locks in your hosting strategy, your hiring profile, your performance ceiling, and your migration cost for the next 3-5 years. I've watched companies spend $150K migrating off a framework they picked in a two-hour meeting. This post exists so you don't end up in that position.

The shift from 2024 to 2026 is also worth noting. Server Components shipped as stable in React 19. Astro introduced Server Islands. Svelte 5 runes replaced the old reactive stores. These aren't incremental upgrades. They change how you think about where code runs and what ships to the browser. If your mental model is still "SPA with an API," you're working with a 2020 playbook.

The 6 Frameworks That Matter (and 4 That Don't)

The survivor list:

  1. Next.js 16 -- The React default. App Router is stable, Server Components are real, Turbopack is the production bundler.
  2. Astro 5 -- Content sites done right. Multi-framework islands, Server Islands, Sessions API, zero JS by default.
  3. SvelteKit -- Svelte 5 runes make reactivity explicit. Smallest bundles in the group. Underrated for production apps.
  4. Nuxt 3 -- Vue 3's meta-framework. Strong in Europe, Asia, and any team already committed to the Vue ecosystem.
  5. Remix (React Router 7) -- Web fundamentals first. Forms, loaders, actions, progressive enhancement. Now merged into React Router.
  6. Gatsby -- Legacy. Still running on thousands of enterprise sites. But Netlify gutted the team, the plugin ecosystem is rotting, and no significant updates since mid-2024.

Who NOT to bet on:

  • Create React App (CRA): Officially dead. React docs no longer recommend it. If you're still on CRA, migrate to Next.js or Vite.
  • Blitz.js: The "Ruby on Rails for React" pitch didn't land. Core team moved on. Community is a ghost town.
  • RedwoodJS: Interesting ideas, abandoned momentum. The 1.0 hype never translated into production adoption at scale.
  • Plain Vite SPAs for content sites: Vite is an excellent build tool, but shipping a client-rendered SPA for a marketing site in 2026 means you're fighting Google's crawler, burning Lighthouse points, and paying a hydration tax your visitors don't deserve.

Framework-by-Framework Deep Dive

Next.js 16

What it is. Next.js is the React meta-framework maintained by Vercel. It's the most widely adopted framework in this list by a factor of 3-4x in npm downloads, job postings, and GitHub activity. If you're building with React in 2026 and you don't have a strong reason to pick something else, you're probably picking Next.js.

What it ships in 2026. Next.js 16 stabilizes the App Router that was introduced in 13 and refined through 14 and 15. Server Components are the default. Server Actions (now called Server Functions in the React docs) let you write mutations that run on the server without building an API route. Turbopack replaces Webpack as the production bundler, cutting build times by 40-60% on large codebases. Partial Prerendering (PPR) lets you serve a static shell with streaming dynamic holes, which is the biggest performance win for authenticated apps since incremental static regeneration.

Real Lighthouse scores. Across our portfolio of 40+ Next.js production sites, mobile Lighthouse performance scores range from 82 to 96. The median is 89. The low end comes from sites with heavy third-party scripts (analytics, chat widgets, A/B testing). A clean Next.js 16 marketing page with optimized images and minimal client JS scores 92-96 consistently.

Hosting cost trajectory. Vercel's Pro plan is $20/seat/month, and you hit the real cost in bandwidth and function invocations. At 100K monthly visitors, expect $40-120/month. At 1M monthly visitors, $180-600/month depending on dynamic vs static ratio. At 10M, $1,200-4,500/month. Cloudflare Pages with the OpenNext adapter is the budget play: $0-50/month at 1M visitors, though you lose some Vercel-specific features. AWS Amplify falls somewhere in between. If you want to go deeper on what we build with Next.js, see our Next.js development capabilities.

Build complexity. A senior React developer can ship a production Next.js 16 site in 2-4 weeks. A junior developer with React experience needs 4-8 weeks and will get tripped up by the Server Component / Client Component boundary. The App Router mental model is genuinely different from the Pages Router, and tutorials from 2023 will actively mislead you.

Hiring market. React is the most common frontend skill in the US market. Senior Next.js developers command $150K-220K/year (US, full-time) or $100-180/hour (contract). The talent pool is deep, but "knows React" does not mean "knows the App Router." Vet for Server Components experience specifically.

When you should pick it. SaaS products, authenticated dashboards, e-commerce with dynamic personalization, any project where your team is already React. It's also the safest bet for enterprise buy-in because every decision-maker has heard of it.

When you shouldn't. Pure content sites where you don't need client interactivity. You'll ship 80-150KB of React runtime to render what could be static HTML. That's where Astro wins.

Migration complexity from WordPress: 3/5. Moderate. The data layer is the hard part. You'll need a headless CMS (Sanity, Contentful, Payload) and a redirect strategy. The build is straightforward, but content modeling takes time. We cover this in detail at WordPress to Next.js migration.

Astro 5

What it is. Astro is the content-first framework that ships zero JavaScript by default. You write components in Astro's own template syntax or bring your own (React, Vue, Svelte, Preact, Solid), and Astro renders them to static HTML at build time. Interactive components hydrate on the client only when you explicitly opt in with client:load, client:visible, or client:idle directives. This "islands architecture" means your blog post page ships 0KB of JS unless you have an interactive widget.

What it ships in 2026. Astro 5 brings Server Islands (render specific components on the server at request time while the rest of the page is static or cached), the Sessions API (first-party session management without external state stores), Content Layer improvements for type-safe Markdown/MDX, and a refined View Transitions API. The Starlight documentation framework is mature and widely adopted.

Real Lighthouse scores. This is where Astro embarrasses everyone. Across 25+ production content sites in our portfolio, mobile Lighthouse performance scores range from 95 to 100. The median is 98. When you ship zero JS, it's hard to score badly. LCP under 1.2s is the norm, not the exception.

Hosting cost trajectory. Static Astro sites on Cloudflare Pages cost $0 at 1M monthly visitors. Free tier. Netlify's free tier handles most content sites. Vercel static hosting is also effectively free at that scale. If you're using Server Islands or SSR mode, costs go up slightly: $5-30/month at 1M visitors on Cloudflare, $30-80/month on Vercel. This is the cheapest framework to host by a wide margin.

Build complexity. A junior developer with basic HTML/CSS/JS skills can ship a production Astro site in 1-2 weeks. Seriously. The learning curve is the gentlest in this list. If you need interactivity, you bring whatever UI library your team knows, so the ceiling is as high as your team's skill.

Hiring market. Astro-specific experience is rarer than React, but you don't need it. Any frontend developer can learn Astro's template syntax in a day. The islands model means you hire for your component library of choice (React, Vue, Svelte) and teach the Astro layer. US senior dev rates: $130K-190K/year or $90-150/hour.

When you should pick it. Marketing sites, blogs, documentation, content-heavy publishers, programmatic SEO at scale (we've built Astro sites with 50,000+ pages that build in under 3 minutes), agency sites where performance is a selling point.

When you shouldn't. Highly interactive SaaS apps, real-time dashboards, anything where most pages require authentication and dynamic data. Astro can do SSR, but it wasn't designed for app-like interactivity, and you'll fight it. For content projects, check out our Astro development capabilities.

Migration complexity from WordPress: 2/5. Easiest migration target for content sites. Export WordPress content to Markdown or connect a headless CMS, map templates to Astro layouts, set up redirects. A 50-page marketing site migrates in 2-4 weeks.

SvelteKit

What it is. SvelteKit is the official meta-framework for Svelte. Svelte compiles your components to vanilla JavaScript at build time, which means no virtual DOM diffing at runtime and dramatically smaller bundles. Svelte 5 introduced "runes," which replace the old $: reactive statements with explicit $state(), $derived(), and $effect() primitives. This was controversial, but in practice, runes make large codebases more predictable and easier to debug.

What it ships in 2026. Svelte 5 runes are stable. SvelteKit's adapter system supports Vercel, Netlify, Cloudflare, Node, and static output. Server-side rendering is a first-class citizen. Form actions handle mutations without client-side state management. The new $effect.tracking() and fine-grained reactivity system means fewer accidental re-renders than any React-based framework.

Real Lighthouse scores. Mobile Lighthouse scores from our SvelteKit production sites range from 90 to 98, median 94. Bundles are typically 30-60% smaller than equivalent Next.js apps because there's no React runtime. A SvelteKit app with a few interactive components often ships 25-40KB of JS total where a Next.js equivalent ships 85-120KB.

Hosting cost trajectory. Similar to Next.js in SSR mode. Vercel: $40-120/month at 100K visitors, $180-500/month at 1M. Cloudflare Pages: $0-40/month at 1M. The smaller bundle sizes mean lower bandwidth costs at scale, saving 10-20% versus Next.js on bandwidth-metered hosts.

Build complexity. If your team knows Svelte, 2-4 weeks to ship. If your team is learning Svelte from React, add 2-3 weeks for the mental model shift. Svelte 5 runes actually make the transition from React easier than old Svelte, because $state() is conceptually closer to useState() than the old implicit reactivity.

Hiring market. This is SvelteKit's biggest weakness. The Svelte talent pool is maybe 10-15% the size of the React pool. US senior rates: $140K-200K/year or $100-160/hour. You'll often hire React developers who want to learn Svelte, which works, but plan for ramp-up time.

When you should pick it. Performance-critical apps where bundle size matters (mobile-first products, emerging markets with slow connections), teams that are already in the Svelte ecosystem, projects where you can invest in developer training.

When you shouldn't. Enterprise projects where you need to backfill developers quickly. Large teams where React experience is the hiring filter. Projects that rely heavily on React-specific libraries (headless UI, Radix, etc.).

Migration complexity from WordPress: 3/5. Similar to Next.js, but fewer off-the-shelf CMS integrations. You'll likely pair with a headless CMS that has a REST or GraphQL API.

Nuxt 3

What it is. Nuxt 3 is to Vue 3 what Next.js is to React. It's the full-featured meta-framework with file-based routing, server-side rendering, static generation, and a rich module ecosystem. If your team writes Vue, Nuxt is your meta-framework. There's no real alternative in the Vue ecosystem.

What it ships in 2026. Nuxt 3 runs on Nitro (a server engine that deploys anywhere) and uses Vite for builds. The Composition API is the standard. Nuxt DevTools are excellent. Server routes and API routes are built in. The module ecosystem (Nuxt Content, Nuxt Image, Nuxt SEO) covers most common needs without third-party packages.

Real Lighthouse scores. Mobile Lighthouse from production: 85-95, median 90. Comparable to Next.js. Vue's runtime is slightly smaller than React's, so bundle sizes trend 10-15% lower at baseline.

Hosting cost trajectory. Nearly identical to Next.js. Vercel supports Nuxt natively. Netlify and Cloudflare Pages work well. $150-500/month at 1M monthly visitors depending on the rendering strategy.

Build complexity. Vue developers ship in 2-4 weeks. React developers learning Vue need 3-5 weeks. The Composition API is closer to React Hooks than the old Options API, so the transition is manageable.

Hiring market. Vue is strong in Europe (especially France, the Netherlands, Germany) and across Asia. In the US, the Vue talent pool is roughly 25-30% the size of React's. US senior rates: $130K-190K/year. If your team or market leans Vue, Nuxt is the obvious choice. If you're starting from scratch in the US, the React/Next.js hiring pool is significantly deeper.

When you should pick it. Your team already writes Vue. You're based in a region where Vue talent is abundant. You prefer Vue's template syntax and Composition API over JSX.

When you shouldn't. You're staffing up a US-based team from zero. You need the absolute largest ecosystem of third-party components and integrations (React wins here by volume).

Migration complexity from WordPress: 3/5. Same as Next.js. Headless CMS + Nuxt frontend + redirect mapping.

Remix (React Router 7)

What it is. Remix merged into React Router in late 2024. React Router 7 with "framework mode" is what we used to call Remix. The philosophy is web fundamentals: HTML forms, HTTP caching, progressive enhancement, and loaders/actions that run on the server. It's opinionated about using the platform instead of reinventing it.

What it ships in 2026. React Router 7 ships with React 19 support, Server Components (partial, opt-in), RSC-based data loading, Vite as the bundler, and first-class Cloudflare Workers support. The loader/action pattern remains the core mental model for data. The framework is leaner than Next.js and gives you more control over the HTTP layer.

Real Lighthouse scores. Mobile Lighthouse: 88-96, median 92. Remix's progressive enhancement approach means pages work without JS, which helps with perceived performance even when Lighthouse numbers are similar to Next.js.

Hosting cost trajectory. Remix runs well on Cloudflare Workers ($5-25/month at 1M visitors), Fly.io ($10-60/month), or any Node.js host. Vercel supports it but it's not the native platform like it is for Next.js. Budget hosts are more viable because Remix doesn't depend on Vercel-specific infrastructure.

Build complexity. React developers familiar with REST patterns ship in 2-4 weeks. The loader/action model is intuitive if you've worked with server-rendered apps before. Developers coming from SPA-only backgrounds may need 1-2 extra weeks to adjust.

Hiring market. Remix-specific talent is niche, but it's React under the hood, so any senior React developer can ramp up quickly. The merging with React Router actually helped because React Router is nearly universal in the React ecosystem.

When you should pick it. Marketplaces, multi-step form-heavy applications, products where URL structure matters deeply, teams that want to stay close to web standards, projects deploying to Cloudflare Workers.

When you shouldn't. Content-heavy sites (Astro is better). Projects where your team expects the Next.js ecosystem of examples, tutorials, and Stack Overflow answers. Enterprise environments where "we use Next.js" is a safer political answer.

Migration complexity from WordPress: 3/5. Similar to Next.js. The data layer is the same challenge.

Gatsby

What it is in 2026. Gatsby is a static site generator built on React and GraphQL. It was the darling of 2019-2021. Netlify acquired it, then effectively abandoned it. The last significant release was mid-2024. The plugin ecosystem is decaying. Core maintainers have moved on. Build times for large sites (10K+ pages) are still painfully slow compared to Astro or Next.js with ISR.

Real Lighthouse scores. Mobile Lighthouse: 75-90, median 82. Gatsby ships a lot of JS by default, and the GraphQL data layer adds overhead that newer frameworks avoid entirely.

Hosting cost trajectory. Gatsby Cloud shut down. You're hosting on Netlify, Vercel, or S3 + CloudFront. Costs are comparable to Next.js static: $50-200/month at 1M visitors.

When you should pick it. You shouldn't. The only reason to touch Gatsby in 2026 is if you inherited a Gatsby site and need to maintain it while planning a migration.

When you shouldn't. Any new project. Period.

Migration complexity from WordPress: N/A. Migrate away from Gatsby, not toward it. If you're on Gatsby and considering your next move, the path is typically Gatsby to Astro for content sites or Gatsby to Next.js for app-like sites.

Comparison Tables

Performance & Developer Experience Matrix

Framework Avg Lighthouse Mobile Avg LCP (s) Avg INP (ms) Bundle Size (KB) Cold Start (ms) Build Time (1K pages)
Next.js 16 89 1.8 120 85-140 150-400 45-90s
Astro 5 98 0.9 40 0-30 50-120 20-40s
SvelteKit 94 1.3 75 25-55 100-250 30-60s
Nuxt 3 90 1.6 110 70-120 140-350 50-100s
Remix (RR7) 92 1.4 90 70-110 80-200 35-70s
Gatsby 82 2.4 180 120-200 N/A (static) 180-600s

Hosting Cost at 1M Monthly Visitors (USD/month)

Framework Vercel Netlify Cloudflare Pages AWS Amplify
Next.js 16 (SSR) $180-600 $200-500 $20-80* $150-400
Astro 5 (Static) $0-20 $0-19 $0 $0-15
Astro 5 (SSR) $30-80 $40-100 $5-30 $30-80
SvelteKit (SSR) $180-500 $180-450 $15-60 $120-350
Nuxt 3 (SSR) $150-500 $180-450 $15-60 $120-350
Remix (SSR) $150-400 $150-400 $5-25 $100-300
Gatsby (Static) $50-200 $50-200 $0-10 $40-150

*Cloudflare Pages with OpenNext adapter; some Vercel-specific features unavailable.

Use-Case Fit

Use Case Next.js 16 Astro 5 SvelteKit Nuxt 3 Remix Gatsby
Marketing site Good Best Good Good Good Avoid
Blog / content hub Good Best Good Good Good Avoid
E-commerce Best Avoid Good Good Good Avoid
SaaS application Best Avoid Good Good Good Avoid
Multi-tenant platform Best Avoid Good Good Best Avoid
Programmatic SEO (50K+ pages) Good Best Good Good Good Avoid

The Decision Tree

Run this in 60 seconds. Go top to bottom, take the first match.

  1. If your team is React engineers building a SaaS app or authenticated product → Next.js 16 with App Router.
  2. If you're building a content site, marketing site, blog, or documentation → Astro 5. No contest on performance or hosting cost.
  3. If you want the smallest bundles and your team is comfortable learning Svelte 5 runes → SvelteKit.
  4. If your team already writes Vue, or you're hiring in a Vue-strong market → Nuxt 3.
  5. If you're building a marketplace, form-heavy product, or deploying to Cloudflare Workers → Remix / React Router 7.
  6. If you're currently on Gatsby → Stop investing. Plan a migration to Astro (content) or Next.js (app).
  7. If you're on WordPress past 100K monthly visitors and performance/security is a concern → Migrate to Next.js + headless CMS or Astro + headless CMS.
  8. If you're a non-technical marketing leader reading this → Hire a team that specializes in headless architecture. The framework choice matters less than the team's experience with it. Start with our headless CMS development page for context.

Real Production Stack Recommendations

I've shipped north of 5,000 sites across agencies, startups, and enterprise clients since 2014. Here are three stacks I'm betting on in 2026, with real cost ranges from production deployments.

Stack A: Content Site / Marketing

Astro 5 + Sanity + Vercel (or Cloudflare Pages) + Resend for transactional email

This is our default for marketing sites, agency builds, and content publishers. Sanity gives editors a real-time collaborative editing experience. Astro renders everything to static HTML. Vercel or Cloudflare serves it at the edge.

  • Monthly cost at 1M MV: $30-180/month (Sanity free tier covers most marketing sites; Vercel static is near-free; Resend is $20/month for transactional email)
  • Build time: 2-4 weeks for a 30-page marketing site
  • Team size: 1 senior developer + 1 designer
  • Performance: Lighthouse 96-100 on every page

This stack is the one we recommend most often for teams coming off WordPress. The editorial experience in Sanity is genuinely better than wp-admin for structured content, and the frontend performance is in a different league. If this sounds like your situation, talk to us about Astro development.

Stack B: SaaS Application

Next.js 16 App Router + Supabase + Stripe + Vercel + PostHog

This is the 2026 version of the "indie SaaS" stack, but it scales to $10M+ ARR. Supabase handles auth, database (Postgres), and real-time subscriptions. Stripe handles payments. PostHog handles analytics and feature flags. Vercel hosts the Next.js app.

  • Monthly cost at 100K MRR: $280-1,800/month total (Vercel $200-800, Supabase $25-300, Stripe 2.9% + $0.30/txn, PostHog $0-450)
  • Build time: 6-12 weeks for MVP
  • Team size: 2-3 developers
  • Performance: Lighthouse 85-94 depending on dashboard complexity

The key insight: don't over-architect at the start. Supabase's row-level security and Postgres do the work of three separate services. Migrate to dedicated infrastructure when you're past $5M ARR, not before. We've built dozens of SaaS products on this stack, documented in our SaaS development work.

Stack C: Multi-Locale Enterprise

Next.js 16 + Sanity + Cloudflare Pages + Algolia

Enterprise marketing sites with 30+ locales, tens of thousands of pages, and search functionality need a different approach. Cloudflare Pages over Vercel here because the global edge network handles multi-region latency better at this scale, and the cost difference is significant: $500-2,000/month on Cloudflare vs $2,000-5,000/month on Vercel for 5M monthly visitors.

  • Monthly cost at 5M MV across 30 locales: $450-5,000/month (Cloudflare $50-400, Sanity $99-799, Algolia $100-1,500, monitoring/CDN/misc $200-2,300)
  • Build time: 12-24 weeks
  • Team size: 3-5 developers + 1 project manager
  • Performance: Lighthouse 88-95

Algolia handles search across all locales with language-specific ranking. Sanity's localization features handle content in a single document with locale-specific fields. Next.js i18n routing maps locale slugs to the correct content.

Migration Realities

If you're moving off WordPress, Drupal, or Sitecore in 2026, here's what 90% of agencies don't tell you before you sign the contract.

Timeline reality:

  • Marketing sites (up to 100 pages): 6-12 weeks, start to live
  • Content-heavy sites (500-5,000 pages): 10-16 weeks
  • E-commerce: 12-24 weeks
  • Enterprise with integrations (CRM, ERP, DAM): 16-32 weeks

The 5 things that break on cutover:

  1. Redirects. WordPress URL patterns (/category/post-name/, /page/2/) don't match your new framework's defaults. Miss a redirect and your organic traffic drops 20-40% overnight. Map every single URL. Use a crawl tool (Screaming Frog, Sitebulb) to build the redirect list before you write a line of code.

  2. Structured data / schema markup. Your WordPress plugins (Yoast, RankMath) generated JSON-LD automatically. Your new site ships with nothing unless you explicitly build it. Article schema, FAQ schema, breadcrumb schema, organization schema: all need manual implementation.

  3. Hreflang tags. If you have multi-locale content, hreflang must be correct on day one. Google takes weeks to re-crawl, and incorrect hreflang causes the wrong locale to rank. Test with the hreflang tag checker before cutover, not after.

  4. Search index timing. Google may take 2-8 weeks to fully re-index your site after a major URL structure change, even with perfect redirects. Plan for a temporary organic traffic dip of 5-15%. It recovers, but your CMO needs to know it's coming.

  5. Image URLs. WordPress stores images in /wp-content/uploads/YYYY/MM/. Your new site uses a CDN with different paths. If any external site or email links to your old image URLs, those are 404s unless you redirect them. And images are often 40-60% of a site's indexed URLs.

Real cost band: $25K-200K depending on page count, integrations, and content complexity. A 30-page marketing site migration runs $25K-50K. A 2,000-page content site with custom post types, WooCommerce, and three language locales runs $80K-200K.

When to phase vs cutover: Phase when you have more than 500 pages or complex integrations. Run the new framework on a subdomain or subdirectory first. Migrate sections incrementally. Cutover when you have fewer than 200 pages and can validate everything in a staging environment in one sprint.

For detailed migration playbooks, see:

FAQ

What is the most modern web framework in 2026?

Astro 5 and Next.js 16 represent the current state of the art in 2026, but they serve different purposes. Astro 5 is the most modern choice for content-driven sites, shipping zero JavaScript by default and achieving Lighthouse scores of 95-100 consistently. Next.js 16 is the most modern choice for interactive applications, with stable Server Components, Server Functions, and Partial Prerendering. "Modern" in 2026 means server-first rendering with selective client hydration, not single-page applications. The old model of shipping a full JavaScript runtime to render a blog post is no longer considered modern practice.

Is Next.js still the best framework in 2026?

Next.js 16 is still the best general-purpose React framework in 2026, but it is not the best framework for every project. It is the top choice for SaaS apps, e-commerce platforms, and authenticated products because of its deep ecosystem, large talent pool, and Vercel's infrastructure. However, for content sites, Astro 5 outperforms Next.js on speed, cost, and simplicity. For projects where bundle size is critical, SvelteKit ships significantly less JavaScript. Next.js is the safest default if you're unsure, but "safest" and "best" are not the same thing.

Should I use Astro or Next.js for a marketing site?

Use Astro 5 for a marketing site. Astro ships zero JavaScript by default, resulting in Lighthouse mobile scores of 95-100 versus Next.js's typical 89-96 for similar content. Hosting costs are dramatically lower: a static Astro site on Cloudflare Pages costs $0/month at 1M monthly visitors, while a Next.js site on Vercel costs $180-600/month. Build complexity is also lower; a junior developer can ship an Astro marketing site in 1-2 weeks. The only reason to pick Next.js for a marketing site is if the same codebase also needs to serve an authenticated app section, and even then, you might be better served by running two separate deployments.

Is SvelteKit production-ready in 2026?

Yes, SvelteKit is fully production-ready in 2026. Svelte 5 runes are stable, the adapter ecosystem covers all major hosting platforms, and companies like Apple, Spotify (internal tools), and The New York Times Wirecutter use Svelte in production. SvelteKit ships the smallest bundles of any major framework, typically 30-60% less JavaScript than equivalent Next.js applications. The primary risk is not technical readiness but hiring: the Svelte talent pool is roughly 10-15% the size of React's. If you can staff the team, SvelteKit is an excellent choice.

What's the best React framework for SEO in 2026?

Next.js 16 is the best React framework for SEO in 2026. It supports static generation, server-side rendering, and incremental static regeneration, all of which produce fully-rendered HTML that search engines can index immediately. The Metadata API provides fine-grained control over title tags, meta descriptions, Open Graph tags, and JSON-LD structured data. Partial Prerendering serves static content instantly while streaming dynamic sections, which helps Core Web Vitals. Remix (React Router 7) is also strong for SEO with its server-rendered approach, but Next.js has more built-in SEO tooling and a larger ecosystem of SEO-focused examples.

How much does it cost to host a Next.js site at 1 million monthly visitors?

Hosting a Next.js site at 1 million monthly visitors costs between $20 and $600 per month depending on the platform and rendering strategy. On Vercel Pro, expect $180-600/month for a mix of static and server-rendered pages. On Cloudflare Pages with the OpenNext adapter, costs drop to $20-80/month, though some Vercel-specific features like image optimization require alternative solutions. On AWS Amplify, $150-400/month is typical. If you run a purely static Next.js export (no SSR), costs on any platform drop to $0-50/month. The biggest cost variable is the ratio of server-rendered to static pages, followed by image bandwidth.

Is Gatsby dead?

Gatsby is not technically dead, but it is effectively abandoned as of 2026. Netlify acquired Gatsby in 2023, shut down Gatsby Cloud, and reduced the core team to minimal maintenance. The last significant feature release was mid-2024. The plugin ecosystem is decaying, with many popular plugins unmaintained and incompatible with current Node.js versions. Build times for large sites remain 3-10x slower than Astro or Next.js. If you're on Gatsby today, plan a migration. For content sites, Astro 5 is the natural successor. For app-like sites, Next.js 16. Do not start a new project on Gatsby.

Can I migrate from WordPress to Astro?

Yes, migrating from WordPress to Astro is one of the most straightforward migration paths available in 2026. For content-focused sites, you export WordPress content to Markdown files or connect Astro to a headless CMS like Sanity, Contentful, or Payload. Astro's Content Collections provide type-safe content handling. A typical 50-page marketing site migrates in 2-4 weeks. The main challenges are the same as any WordPress migration: redirect mapping, structured data reimplementation, and image URL handling. The result is a site that scores 95-100 on Lighthouse, costs near-zero to host, and is dramatically more secure than WordPress because there's no server-side runtime to attack. We rate this migration a 2 out of 5 in difficulty for content sites. See our full guide on WordPress alternatives for a detailed comparison.