Resources/Benchmark
BENCHMARKS / PERFORMANCE & CAPABILITY

Query faster. Read accurately.Give agents the power to finish.

How fast can you query a terabyte? How well can you parse complex documents, retrieve relevant results and complete real agent tasks? See the workloads and results.

Reports and public sources · Aug–Oct 2026Conditions includedExplore the results

Data foundation

Branching, JSON, OLTP, HTAP, analytics, time series and logs.

Selected scope · Git4Data parallel branching
MatrixOneMeasured results/Git4Data parallel branching

120x more data, 1.1–1.6x more branch time.

1.1–1.6×
Runtime growth at 120x data

Same branching workflows, data scaled from 250K rows to SF100 (about 120x): MatrixOne grows 1.1–1.6x, DoltDB 68–120x. software_dev (physical ADD COLUMN) did not complete at any tier this round — see the note.

data_cleaning · 10 workers · MatrixOne
37.5 s
Same point · DoltDB
511.6 s
Test scope

SF100 · order_line 30M rows · 10 / 100 workers

Workflows completed by both · completion times · Lower is better
data_cleaning · 10 workers13.6× faster vs DoltDB
MatrixOne
37.5
DoltDB
511.6
data_cleaning · 100 workers4.1× faster vs DoltDB
MatrixOne
560.4
DoltDB
2,275.7
mcts · 10 workers5.0× faster vs DoltDB
MatrixOne
179.0
DoltDB
887.5
mcts · 100 workers1.6× faster vs DoltDB
MatrixOne
3,169.9
DoltDB
5,016.1
05,016.1 s

Only the four points where both systems completed are compared. software_dev performs a physical ALTER TABLE ADD COLUMN per step; MatrixOne did not complete it at 10 / 100 / 500 workers, while DoltDB completed 10 / 100. At 250K rows DoltDB was 7.2–33.4x faster, so the two curves cross in between. Speedups vary with scale and do not form a cross-scale ranking.

Full values & methodology

Workflows completed by both · completion time

SubjectMatrixOne (s)DoltDB (s)
data_cleaning · 10 workers37.5511.6
data_cleaning · 100 workers560.42,275.7
mcts · 10 workers179.0887.5
mcts · 100 workers3,169.95,016.1

Conditions & definitions

Git4Data parallel branching at SF100 (order_line 30,000,000 rows, warehouse 100), 10 / 100 / 500 workers; MatrixOne 77dbf546, DoltDB dolt-sql-server:2.2.3; 2026-10-03 delivery. Comparison point at 250,000 rows. Single run per configuration.

Git4Data · SF100 · completed by both
WorkloadWritersMatrixOne (s)DoltDB (s)SpeedupStatus
data_cleaning1037.5511.613.7×completed
data_cleaning100560.42,275.74.1×completed
mcts10179.0887.55.0×completed
mcts1003,169.95,016.11.6×completed
software_dev10n/a235.3—not completed
software_dev100n/a2,154.1—not completed

SourcesP0-2 Git4Data · 2026-10-03

Studies in this layer

Select a study to explore its results and methodology.

Bring your workload to the conversation.

Evaluate the right approach using your enterprise data, documents and agent tasks.

Discuss your use case