TechEmpower Web Framework Benchmarks

The TechEmpower Web Framework Benchmarks are a long-running, open-source project that compares hundreds of web frameworks and platforms across many languages on identical, standardized workloads. The results, published in periodic rounds, are a widely referenced way to gauge the raw efficiency ceiling of frameworks for Java, .NET, Go, Rust, Node, Python, PHP, and more. They are notable for breadth: few other efforts test so many stacks under the same rules on the same hardware.

What It Measures

The suite defines several test types run against each framework: JSON serialization, single database query, multiple database queries, fortunes (server-side templating with database reads), data updates, and plaintext. The headline metric is requests per second under high concurrency, with latency distributions reported alongside. Together these isolate serialization speed, database access efficiency, ORM overhead, and routing cost, letting readers see where a framework spends its time rather than just an aggregate number.

Methodology

Every framework implements the same specified endpoints with the same database schema, and all are tested on identical hardware (physical and cloud variants) using the wrk load generator at controlled concurrency levels. Contributions come from framework authors and the community, so implementations reflect idiomatic best-practice use, and peer pressure tends to keep them well optimized. Each round documents hardware, database engine (PostgreSQL, MySQL), and configuration. Results are published with full source so anyone can reproduce, audit, or contest them, which keeps the project honest.

How to Interpret Results

Higher requests per second is better, but choose the test type that matches your workload: plaintext and JSON show framework-and-runtime overhead, while the database tests reflect realistic data-driven services. Compare frameworks within the same test type and round, never across rounds, since hardware and software change. Top performers often expose the ceiling of a runtime rather than typical application throughput, so read the database-backed tests for a more realistic signal. Latency percentiles matter as much as throughput for user-facing services, and a framework that leads on raw throughput may not lead on tail latency.

Limitations

The benchmarks measure micro-endpoints, not full applications with business logic, authentication, caching, and complex queries, so leaderboard rank rarely predicts production performance. Implementation quality varies by contributor, and some entries are heavily optimized beyond normal use, occasionally bordering on the unrealistic. Database tests remain relatively simple. Use the suite to understand relative framework efficiency and to shortlist candidates, then validate the finalists with your own representative workload, data volumes, and access patterns.