Kernels

Kernel Catalog

Every package on main, with its reference oracle, its Mojo kernel, and its measured warm speedup. Deep dives exist for the three flagship kernels: bm25 · Fuse · cclib GaussGrid. Methodology and lowlights: Benchmarks.

Python (22 packages — pip install <name>)

Package	Accelerates	Domain	Kernel	Speedup (warm)
bm25-mojo	rank_bm25	text retrieval	kernels/bm25	116–8,769×
cclib-mojo	cclib method.volume	computational chemistry	kernels/gaussgrid	72–7,033×
pypdf-filters-mojo	pypdf stream filters	PDF	kernels/pypdf-filters	12–1,776×
pykalman-mojo	pykalman	state-space models	kernels/pykalman	3.9–827×
elephant-mojo	elephant surrogates	neuroscience	kernels/elephant-surrogates	2–342×
uproot-mojo	uproot	high-energy physics	kernels/uproot-branches	0.9–172.6×
metpy-mojo	MetPy cape_cin	meteorology	kernels/metpy-cape	78–156.6×
dynesty-mojo	dynesty	Bayesian inference	kernels/dynesty	4.3–51.9×
jsonpath-mojo	jsonpath_ng	JSON query	kernels/jsonpath	1.7–49.2×
croniter-mojo	croniter	scheduling	kernels/croniter	15.3–46×
obspy-mojo	ObsPy Konno–Ohmachi	seismology	kernels/obspy-konno	3.9–45×
nx-mojo	networkx	graph analysis	kernels/networkx-graph	1.2–40×
ruptures-mojo	ruptures	change-point detection	kernels/ruptures	2.2–48×
ase-mojo	ASE neighbor list	atomistic simulation	kernels/ase_neighborlist	20.3–28.3×
jsonschema-mojo	jsonschema	validation	kernels/jsonschema	2.1–28×
sacrebleu-mojo	sacrebleu	MT evaluation	kernels/sacrebleu	2.9–15.2×
toml-mojo	tomllib / tomlkit	config parsing	kernels/toml	1.8–8×
langdetect-mojo	langdetect	NLP	kernels/langdetect	4.3–6.9×
vader-mojo	vaderSentiment	sentiment	kernels/vader	4.1×
bio-mojo	Biopython Bio.SeqIO	bioinformatics	kernels/bioparse	1.4–2.3×
ta-mojo	pandas-ta-classic	technical analysis	kernels/ta	1.8–93×
jmespath-mojo	jmespath	JSON query	kernels/jmespath	0.01–1.1× (honest)

TypeScript / Node (4 packages — npm install @<name>-mojo/core)

Package	Accelerates	Domain	Kernel	Speedup (warm)
fuse-mojo (@fuse-mojo/core)	Fuse.js 7.1.0	fuzzy search	kernels/fuse	8.8–38.7×
natural-mojo (@natural-mojo/core)	natural LevenshteinDistance, DamerauLevenshteinDistance	NLP	kernels/natural	3.7–64.5×
ckmeans-mojo (@ckmeans-mojo/core)	simple-statistics ckmeans	clustering	kernels/ckmeans	1.6–3.4×
minisearch-mojo (@minisearch-mojo/core)	MiniSearch fuzzy/prefix/suggest	search	kernels/minisearch_fuzzy	0.8–3.1×

Reading this table

• Ranges span workload cells, not run-to-run noise — small inputs and cold starts are where the low ends live. Peak-cell context is in Benchmarks.
• Every package also ships a vendored pure-language fallback (asserted at the same tolerance), per-platform self-contained binaries, an <NAME>_MOJO_DISABLE_NATIVE=1 escape hatch, and a backend_info() diagnostic. See Getting Started.
• Missing your library? The target-selection bar and the contribution path are in Writing a Kernel; requests go to GitHub Issues.

---

Apache-2.0, © 2026 Algenta