Proof · 01 — The receipts

Real numbers. Sourced. Re-run on every commit.

Every figure on this page comes from a benchmark file in the engine repo, with the source named below it. They re-run on every commit; the page is updated when the numbers move. The disclosures at the bottom are part of the page, not buried.

191/192: tests passing
22.9ms: schema compile · 1k models
2.4µs: SELECT compile
88.97KB: ESM bundle

v0.1.0 · earlynumbers reflect engine compilation overhead — not end-to-end query timelast run · CI · main

02 — Schema-scale benchmarksscale.test.ts

How the engine behaves at a thousand models.

row · 0110 / 100 / 500 / 1k

22.9ms

schema compile · 1000 models

Lex + parse + validate the .zanith source into a runtime graph. Sub-linear at scale because validation reuses tokenization state.

src · engine/test/benchmark/scale.test.ts

row · 0210 / 100 / 500 / 1k

0.73ms

1000 model lookups

Indexed lookup against the in-memory model registry. Sub-microsecond per call — well below any human-perceptible threshold.

src · engine/test/benchmark/scale.test.ts

row · 0310 / 100 / 500 / 1k

3.4MB

graph memory · 1000 models

Total memory footprint of the runtime graph at 1k models. Order of magnitude smaller than a comparably-sized generated client on disk.

src · engine/test/benchmark/scale.test.ts

03 — Per-operation costexecution.test.ts · 10,000 iterations each

Every measured operation, all on one page.

groupoperationcost / op

expression

Expression · simple eq

{ field: value }

1.2µs

expression

Expression · AND/OR

{ AND: [a, b], OR: [c, d] }

2.4µs

compile

SELECT · with WHERE

where + projection

2.4µs

compile

findMany · build + compile

args validate + AST + emit

2.1µs

compile

JOIN · projection · WHERE

1 included relation

17.2µs

compile

GROUP BY + COUNT + SUM

aggregate compile

5.5µs

write

INSERT · single row

RETURNING *

2.8µs

write

UPSERT · ON CONFLICT DO UPDATE

deduped by unique field

2.7µs

write

Bulk INSERT · 10 rows

values list

7.6µs

9 operations · 4 groupseach = engine overhead, not query execution time

04 — The test suite191 / 192 · disclosed

191 of 192 passing. One flaky benchmark, called out plainly.

vitest · run #1247main · 19 files

$ pnpm test
 
vitest run · zanith engine
 
✓ test/compiler/select.test.ts (8)
✓ test/compiler/insert.test.ts (5)
✓ test/compiler/update.test.ts (4)
✓ test/compiler/delete.test.ts (3)
✓ test/expression/expr.test.ts (12)
✓ test/expression/where.test.ts (18)
✓ test/integration/pipeline.test.ts (6)
✓ test/integration/typed-client.test.ts (9)
✓ test/edge-cases/negative.test.ts (22)
✓ test/edge-cases/null-handling.test.ts (8)
✓ test/types/end-to-end.test.ts (13)
✓ test/types/inference.test.ts (7)
✓ test/schema/parser.test.ts (15)
✓ test/schema/validator.test.ts (11)
✓ test/schema/relations.test.ts (9)
✓ test/schema/builder.test.ts (12)
✓ test/benchmark/scale.test.ts (6) 427ms
⚠ test/benchmark/execution.test.ts (1) 466ms
  ❯ pipeline-stages: ratio assertion (flaky)
 
  Test Files  19 passed (19)
       Tests  191 passed | 1 failed (192)
    Duration  1.03s

committed at every push · ci-green policy on maindisclosed honestly, not rounded

The 192nd test, in plain English

A ratio asserting against a near-zero baseline.

The flaky assertion compares two timings — a relational query versus a near-zero raw-string-concat baseline — and divides them. Because the baseline is so close to zero, tiny CPU jitter swings the ratio by 100x between runs. The per-op check directly above it (< 50µs) is stable and passes every time. We're reporting 191 / 192 rather than rounding to all-green; the flaky check will be rewritten to drop the ratio.

honest disclosure

05 — What this page doesn't measure

The numbers above are real. Here's what they aren't.

Per the VOICE.md disclaimer rule: every claim made on this page has an accompanying caveat about what the figure does and does not cover. Four gaps, listed plainly so the credibility is shaped by the absence too.

Gap · 01scope · by design
End-to-end query time against a live database
Every µs on this page is engine overhead. Network latency, connection setup, and the database's actual execution time are excluded. If you want end-to-end numbers, the local benchmark setup in §06 will produce them.
Gap · 02in progress
Comparable benchmarks against Prisma, Drizzle, TypeORM
We don't have sourced competitor figures yet. Until we do, the marketing copy uses one consistent hedge: "codegen ORMs at this scale spend minutes regenerating." Specific µs comparisons will land when we can run a matched workload on each.
Gap · 03planned
Long-running memory churn
The 3.4MB figure is the steady-state graph footprint. We haven't yet measured what happens to memory under continuous schema reloads, sustained query throughput, or under-leak-detection runs over hours.
Gap · 04downstream concern
Performance on hot-paths inside the DB
The compiler's choice of joins, projections, and parameter shapes affects how the database plans the query. We measure what we emit, not how the planner reacts. The /examples page shows the SQL we generate; production tuning lives downstream.

Ultimate Trust

Reproducible benchmarks

Benchmark Script

~/zanith/engine $ git clone <engine-repo>
~/zanith/engine $ cd packages/benchmarks
~/zanith/engine/benchmarks $ npm install

# Run the 1000-model massive graph benchmark
~/zanith/engine/benchmarks $ npx tsx run.ts --type massive

✓ Generating 1000 interconnected models... (done)
✓ Compiling runtime graph...
> Zanith Core Initialized in 59.12ms
> Peak RSS Memory: 14.3MB

"Don't trust this page. Run it yourself."

07 — Where to go from here

door · 01

Real numbers. Sourced. Re-run on every commit.

How the engine behaves at a thousand models.

Every measured operation, all on one page.

191 of 192 passing. One flaky benchmark, called out plainly.

A ratio asserting against a near-zero baseline.

The numbers above are real. Here's what they aren't.

End-to-end query time against a live database

Comparable benchmarks against Prisma, Drizzle, TypeORM

Long-running memory churn

Performance on hot-paths inside the DB

Reproducible benchmarks

How it works

Examples

Roadmap