Benchmark

Our open-source benchmark measures pure SQL generation speed, with no database I/O, to isolate the overhead each tool adds to every query. This page always reflects the latest run; the original announcement post tells the full story.

Node.js v24, Apple Silicon M-series, July 2026. Latest stable of every entry: TypeORM 1.0.0, MikroORM 7.1.5, Sequelize 6.37.8, Drizzle 0.45.2, Knex 3.3.0, Kysely 0.29.3, UQL 0.15.5. All values in ops/sec (higher is better), averaged over 3 runs.

Results

Query Type	UQL	Sequelize	TypeORM	MikroORM	Drizzle	Knex	Kysely
INSERT (10 rows)	472K	205K	43K	108K	11K	464K	187K
UPDATE (SET+WHERE)	1,748K	237K	294K	210K	74K	699K	729K
UPSERT (ON CONFLICT)	591K	343K	266K	259K	34K	451K	314K
DELETE (WHERE)	3,388K	1,233K	519K	250K	195K	1,093K	1,177K
SELECT (1 field)	3,951K	3,154K	605K	565K	217K	1,117K	1,456K
SELECT (WHERE+SORT+LIMIT)	1,299K	396K	285K	74K	56K	607K	362K
SELECT (complex $or)	663K	149K	167K	28K	32K	242K	195K
AGGREGATE (GROUP+HAVING)	1,494K	404K	280K	73K	67K	326K	187K

UQL is the fastest entry in all 8 categories, on average ~2.1x faster than the next fastest tool. Seven of those wins are clear, ranging from 1.2x to 3.7x. The eighth, batch INSERT, is the narrowest: UQL edges the standalone query builder Knex by a few percent (472K vs. 464K here) and stays ahead of it across every run, rather than trading the lead run to run as earlier versions did. Explore the interactive charts.

Operation group	Avg advantage (vs. next fastest)
Reads (SELECT, filter, complex, AGGREGATE)	~2.4x
Writes (UPDATE/UPSERT/DELETE)	~2.1x

Tip

Query Builder vs. ORM: The benchmark includes Knex and Kysely (pure query builders) to show that UQL keeps pace with tools that carry no entity or relation overhead. It is faster in every category, most by a wide margin and batch INSERT by a slim one.

Why UQL is fast

Two design choices drive the numbers:

• Schema metadata (tables, columns, relations) is pre-computed once at startup, so no lookups happen at query time.
• SQL is written directly into a string buffer, avoiding intermediate builder objects and garbage-collection pressure.

Methodology

• What’s measured: pure SQL string generation, with no database, no I/O, and no connection pool.
• Fairness: every entry defines the same User entity, generates logically equivalent queries, compiles the PostgreSQL dialect, and uses its idiomatic API (QueryBuilder for TypeORM/MikroORM, QueryGenerator for Sequelize).
• Why no Prisma? Prisma’s query compiler is Rust compiled to WebAssembly, not pure JS/TS, and it exposes no public SQL-compilation API, so it cannot be measured this way.
• Reproduce it: no database needed; it runs in seconds.

git clone https://github.com/rogerpadilla/ts-orm-benchmark.git
cd ts-orm-benchmark
bun install
npm run bench

The full methodology, fairness guarantees, and per-run data live in the benchmark repository.

---

Speed is one axis. For side-by-side API and feature differences, see the ORM Comparison.

ORM Comparison →UQL vs Prisma, Drizzle, TypeORM, and MikroORM, with actual code for every common operation.

---