Chart · millijoules a run
filter amount > 500, count, in energy.
neo 1.08 mJ in the energy run. Closest: Polars at 1.92×. Farthest: Spark at 1,313×.
Embed anywhere.
<iframe src="https://n0.is/embed/chart/filter-energy" title="filter amount > 500, count, in energy" width="100%" height="520" loading="lazy" sandbox="allow-scripts" referrerpolicy="no-referrer" style="border:0;border-radius:12px"></iframe>Every engine's multiple of neo
- neo1×
- Polars1.92× · 2.07 vs 1.08
- pandas2.42× · 2.61 vs 1.08
- DuckDB13.7× · 14.8 vs 1.08
- Spark1,313× · 1,413 vs 1.08
Bars to scale, from zero to the slowest engine. The gray pair is the engine's measurement and neo's in the same run, in mJ.
What this measures
Count the rows whose amount is over 500: a million comparisons on a typed column, the answer one integer. Each library's count is written the way its users write it. Energy is the same run in millijoules: what the machine billed each engine's whole process tree, with its draw at rest subtracted for the seconds of the window. neo's pool is told to rest after the kernel, as the editor tells it.
What neo is held to
The count every other engine prints, under the JavaScript node's rule that NaN never matches. What comes out is a selection the next kernel reads from, not a copy of the rows.
Each side, this kernel only
Polars1.92× neo · neo vs Polars
timed("filter", lambda: df.lazy().filter(pred).select(pl.len()).collect().item())pandas2.42× neo · neo vs pandas
timed("filter", lambda: int((amount > 500).sum()))DuckDB13.7× neo · neo vs DuckDB
timed("filter", lambda: con.execute("SELECT count(*) FROM t WHERE amount > 500").fetchone())Spark1,313× neo · neo vs Spark
timed("filter", lambda: df.filter(pred).count())neo · confidential
neo's kernels, the ports written in-house to beat them, and the techniques behind both stay with the engine. Nothing about how the numbers were made is withheld: the method, the checks every engine had to pass, every number and every place neo is behind are on this page. The engine itself we keep confidential, because we don't want it getting into the wrong hands. The binaries run on request and reproduce every number here.
Measured on Apple M5 Max, 18 cores, 26.4.1, 2026-10-11. Every number here is also on the comparison page of its engine, with how it was taken and the caveats.