Concordance-gated public results

Benchmarks

AutoZyme reports wall-clock speedups against upstream toolkits, with method-specific output checks for every run. The first view is grouped by accelerated task; the raw run table remains below for audit.

Public runs

Accelerated tasks

8.5×

Median passing speedup

4,363×

Peak passing speedup

98.0%

96/98 pass

Fastest win
4,363×
Scanpy sc_rank_genes
heart_adult · 32 threads · 21.81 min to 300 ms

Seurat path
69.9×
run_pca
pbmc68k · 1.86 min to 1.60 s

Longest baseline
86.29 min
Seurat sctransform
Optimized to 1.80 min with 47.9× speedup

Coverage
18 tasks
2 toolkits · 5 datasets
96/98 public runs pass their concordance gate

Accelerated tasks

18 of 18 tasks shown

Preprocessing

7 tasks

Seurat5/5 pass

normalize_data

Best59.5×

Median2.00×

Datasets3

pbmc68k · 32 threads2.38 s to 40 ms

heart_adult, pbmc200k, pbmc68k1, 32 threadsmin concordance 1.000 data_cor

Dataset	Threads	Baseline	Optimized	Speedup	Concordance	Pass
pbmc68k	32	2.38 s	40 ms	59.5×	1.000	pass
pbmc200k	32	12.79 s	270 ms	47.4×	1.000	pass
pbmc68k	1	2.38 s	1.19 s	2.00×	1.000	pass
pbmc200k	1	12.79 s	7.17 s	1.78×	1.000	pass
heart_adult	1	29.10 s	18.38 s	1.58×	1.000	pass

Seurat3/3 pass

sctransform

Best47.9×

Median18.1×

Datasets2

pbmc200k · 1 threads86.29 min to 1.80 min

pbmc200k, pbmc68k1, 32 threadsmin concordance 1.000 residual_cor

Dataset	Threads	Baseline	Optimized	Speedup	Concordance	Pass
pbmc200k	1	86.29 min	1.80 min	47.9×	1.000	pass
pbmc68k	32	3.81 min	12.67 s	18.1×	1.000	pass
pbmc68k	1	3.81 min	24.41 s	9.38×	1.000	pass

Scanpy6/6 pass

sc_highly_variable

Best34.5×

Median16.3×

Datasets3

heart_adult · 32 threads6.90 s to 200 ms

heart_adult, pbmc200k_glaucoma, pbmc68k1, 32 threadsmin concordance 1.000 hvg_overlap

Dataset	Threads	Baseline	Optimized	Speedup	Concordance	Pass
heart_adult	32	6.90 s	200 ms	34.5×	1.000	pass
heart_adult	1	3.50 s	200 ms	17.5×	1.000	pass
pbmc200k_glaucoma	1	1.50 s	100 ms	15.0×	1.000	pass
pbmc200k_glaucoma	32	1.50 s	100 ms	15.0×	1.000	pass
pbmc68k	1	200 ms	0 ms	—	1.000	pass
pbmc68k	32	200 ms	0 ms	—	1.000	pass

Scanpy6/6 pass

sc_normalize

Best31.7×

Median20.8×

Datasets3

heart_adult · 32 threads9.50 s to 300 ms

heart_adult, pbmc200k_glaucoma, pbmc68k1, 32 threadsmin concordance 1.000 data_cor

Dataset	Threads	Baseline	Optimized	Speedup	Concordance	Pass
heart_adult	32	9.50 s	300 ms	31.7×	1.000	pass
heart_adult	1	4.50 s	200 ms	22.5×	1.000	pass
pbmc200k_glaucoma	1	1.90 s	100 ms	19.0×	1.000	pass
pbmc200k_glaucoma	32	1.90 s	100 ms	19.0×	1.000	pass
pbmc68k	1	300 ms	0 ms	—	1.000	pass
pbmc68k	32	300 ms	0 ms	—	1.000	pass

Seurat5/5 pass

scale_data

Best18.9×

Median9.04×

Datasets3

pbmc68k · 32 threads2.08 s to 110 ms

heart_adult, pbmc200k, pbmc68k1, 32 threadsmin concordance 1.000 data_cor

Dataset	Threads	Baseline	Optimized	Speedup	Concordance	Pass
pbmc68k	32	2.08 s	110 ms	18.9×	1.000	pass
pbmc200k	32	7.80 s	530 ms	14.7×	1.000	pass
pbmc68k	1	2.08 s	230 ms	9.04×	1.000	pass
pbmc200k	1	7.80 s	1.18 s	6.61×	1.000	pass
heart_adult	1	16.63 s	2.62 s	6.35×	1.000	pass

Seurat5/5 pass

find_variable_features

Best16.8×

Median8.07×

Datasets3

pbmc200k · 32 threads10.89 s to 650 ms

heart_adult, pbmc200k, pbmc68k1, 32 threadsmin concordance 1.000 top2000_jaccard

Dataset	Threads	Baseline	Optimized	Speedup	Concordance	Pass
pbmc200k	32	10.89 s	650 ms	16.8×	1.000	pass
pbmc200k	1	10.89 s	1.19 s	9.15×	1.000	pass
pbmc68k	32	2.26 s	280 ms	8.07×	1.000	pass
pbmc68k	1	2.26 s	350 ms	6.46×	1.000	pass
heart_adult	1	22.51 s	11.93 s	1.89×	1.000	pass

Scanpy6/6 pass

sc_scale

Best8.01×

Median5.46×

Datasets3

pbmc68k · 1 threads500 ms to 62 ms

heart_adult, pbmc200k_glaucoma, pbmc68k1, 32 threadsmin concordance 1.000 data_cor

Dataset	Threads	Baseline	Optimized	Speedup	Concordance	Pass
pbmc68k	1	500 ms	62 ms	8.01×	1.000	pass
pbmc68k	32	500 ms	63 ms	7.90×	1.000	pass
heart_adult	32	4.30 s	743 ms	5.79×	1.000	pass
heart_adult	1	2.80 s	546 ms	5.13×	1.000	pass
pbmc200k_glaucoma	32	1.40 s	273 ms	5.13×	1.000	pass
pbmc200k_glaucoma	1	1.40 s	277 ms	5.05×	1.000	pass

Dimensionality reduction

4 tasks

Seurat6/6 pass

run_pca

Best69.9×

Median30.0×

Datasets3

pbmc68k · 32 threads1.86 min to 1.60 s

heart_adult, pbmc200k, pbmc68k1, 32 threadsmin concordance 1.000 embedding_cor

Dataset	Threads	Baseline	Optimized	Speedup	Concordance	Pass
pbmc68k	32	1.86 min	1.60 s	69.9×	1.000	pass
pbmc200k	32	2.72 min	3.08 s	52.9×	1.000	pass
heart_adult	32	3.87 min	6.62 s	35.1×	1.000	pass
pbmc68k	1	1.86 min	4.47 s	25.0×	1.000	pass
pbmc200k	1	2.72 min	12.99 s	12.5×	1.000	pass
heart_adult	1	3.34 min	28.64 s	7.01×	1.000	pass

Scanpy6/6 pass

sc_pca

Best28.7×

Median16.0×

Datasets3

pbmc200k_glaucoma · 32 threads57.40 s to 2.00 s

heart_adult, pbmc200k_glaucoma, pbmc68k1, 32 threadsmin concordance 1.000 embedding_cor

Dataset	Threads	Baseline	Optimized	Speedup	Concordance	Pass
pbmc200k_glaucoma	32	57.40 s	2.00 s	28.7×	1.000	pass
pbmc68k	32	21.50 s	900 ms	23.9×	1.000	pass
heart_adult	32	1.35 min	4.80 s	16.9×	1.000	pass
pbmc68k	1	33.20 s	2.20 s	15.1×	1.000	pass
pbmc200k_glaucoma	1	1.45 min	6.90 s	12.6×	1.000	pass
heart_adult	1	1.65 min	14.90 s	6.65×	1.000	pass

Seurat6/6 pass

run_umap

Best3.79×

Median3.35×

Datasets3

pbmc200k · 1 threads4.54 min to 1.20 min

heart_adult, pbmc200k, pbmc68k1, 32 threadsmin concordance 0.963 embedding_cor

Dataset	Threads	Baseline	Optimized	Speedup	Concordance	Pass
pbmc200k	1	4.54 min	1.20 min	3.79×	0.998	pass
pbmc68k	32	1.51 min	26.38 s	3.44×	0.963	pass
pbmc68k	1	1.51 min	26.90 s	3.37×	0.963	pass
heart_adult	32	11.02 min	3.31 min	3.33×	0.997	pass
pbmc200k	32	4.54 min	1.46 min	3.11×	0.998	pass
heart_adult	1	13.00 min	4.71 min	2.76×	0.997	pass

Scanpy6/6 pass

sc_umap

Best2.00×

Median1.58×

Datasets3

heart_adult · 1 threads4.61 min to 2.31 min

heart_adult, pbmc200k_glaucoma, pbmc68k1, 32 threadsmin concordance 0.971 embedding_cor

Dataset	Threads	Baseline	Optimized	Speedup	Concordance	Pass
heart_adult	1	4.61 min	2.31 min	2.00×	1.000	pass
pbmc68k	1	36.00 s	21.50 s	1.67×	0.971	pass
pbmc200k_glaucoma	1	1.65 min	1.02 min	1.61×	0.999	pass
pbmc68k	32	33.60 s	21.60 s	1.56×	1.000	pass
pbmc200k_glaucoma	32	1.60 min	1.05 min	1.53×	0.999	pass
heart_adult	32	4.41 min	2.98 min	1.48×	1.000	pass

Graph + clustering

4 tasks

Scanpy6/6 pass

sc_neighbors

Best35.6×

Median9.96×

Datasets3

pbmc68k · 32 threads32.00 s to 900 ms

heart_adult, pbmc200k_glaucoma, pbmc68k1, 32 threadsmin concordance 0.917 nn_overlap

Dataset	Threads	Baseline	Optimized	Speedup	Concordance	Pass
pbmc68k	32	32.00 s	900 ms	35.6×	0.989	pass
pbmc68k	1	30.90 s	1.00 s	30.9×	0.989	pass
pbmc200k_glaucoma	32	40.20 s	4.00 s	10.1×	0.917	pass
pbmc200k_glaucoma	1	38.50 s	3.90 s	9.87×	0.917	pass
heart_adult	32	1.70 min	51.00 s	2.00×	0.925	pass
heart_adult	1	53.50 s	42.80 s	1.25×	0.925	pass

Seurat4/6 pass

find_clusters_leiden

Best8.95×

Median7.07×

Datasets3

heart_adult · 1 threads3.27 min to 21.92 s

heart_adult, pbmc200k, pbmc68k1, 32 threadsmin concordance 0.922 ari

Dataset	Threads	Baseline	Optimized	Speedup	Concordance	Pass
heart_adult	1	3.27 min	21.92 s	8.95×	0.958	pass
heart_adult	32	3.74 min	26.32 s	8.51×	0.958	pass
pbmc68k	1	12.02 s	2.13 s	5.64×	0.990	pass
pbmc200k	1	46.46 s	8.31 s	5.59×	0.922	flag
pbmc68k	32	12.02 s	2.25 s	5.34×	0.990	pass
pbmc200k	32	46.46 s	9.25 s	5.02×	0.922	flag

Scanpy6/6 pass

sc_leiden

Best8.53×

Median7.49×

Datasets3

heart_adult · 1 threads1.96 min to 13.80 s

heart_adult, pbmc200k_glaucoma, pbmc68k1, 32 threadsmin concordance 0.980 ari

Dataset	Threads	Baseline	Optimized	Speedup	Concordance	Pass
heart_adult	1	1.96 min	13.80 s	8.53×	0.980	pass
heart_adult	32	3.20 min	22.90 s	8.38×	0.980	pass
pbmc200k_glaucoma	1	45.60 s	6.00 s	7.60×	0.988	pass
pbmc200k_glaucoma	32	46.50 s	6.30 s	7.38×	0.988	pass
pbmc68k	1	4.30 s	1.00 s	4.30×	0.989	pass
pbmc68k	32	4.20 s	1.00 s	4.20×	0.989	pass

Seurat6/6 pass

find_clusters_louvain

Best3.90×

Median2.83×

Datasets3

pbmc200k · 1 threads1.18 min to 18.23 s

heart_adult, pbmc200k, pbmc68k1, 32 threadsmin concordance 0.961 ari

Dataset	Threads	Baseline	Optimized	Speedup	Concordance	Pass
pbmc200k	1	1.18 min	18.23 s	3.90×	0.977	pass
pbmc200k	32	1.18 min	19.50 s	3.64×	0.977	pass
pbmc68k	1	13.28 s	4.59 s	2.89×	0.974	pass
pbmc68k	32	13.28 s	4.79 s	2.77×	0.974	pass
heart_adult	1	5.28 min	2.23 min	2.37×	0.961	pass
heart_adult	32	6.10 min	2.58 min	2.36×	0.961	pass

Differential expression

2 tasks

Scanpy6/6 pass

sc_rank_genes

Best4,363×

Median1,699×

Datasets3

heart_adult · 32 threads21.81 min to 300 ms

heart_adult, pbmc200k_glaucoma, pbmc68k1, 32 threadsmin concordance 1.000 top100_overlap

Dataset	Threads	Baseline	Optimized	Speedup	Concordance	Pass
heart_adult	32	21.81 min	300 ms	4,363×	1.000	pass
heart_adult	1	13.10 min	300 ms	2,620×	1.000	pass
pbmc200k_glaucoma	1	2.84 min	100 ms	1,704×	1.000	pass
pbmc200k_glaucoma	32	2.82 min	100 ms	1,694×	1.000	pass
pbmc68k	32	12.30 s	100 ms	123.0×	1.000	pass
pbmc68k	1	12.20 s	100 ms	122.0×	1.000	pass

Seurat6/6 pass

find_all_markers

Best48.5×

Median31.9×

Datasets3

heart_adult · 1 threads44.46 min to 54.97 s

heart_adult, pbmc200k, pbmc68k1, 32 threadsmin concordance 1.000 gene_jaccard

Dataset	Threads	Baseline	Optimized	Speedup	Concordance	Pass
heart_adult	1	44.46 min	54.97 s	48.5×	1.000	pass
heart_adult	32	44.45 min	1.03 min	43.3×	1.000	pass
pbmc200k	32	11.44 min	21.35 s	32.2×	1.000	pass
pbmc200k	1	11.44 min	21.64 s	31.7×	1.000	pass
pbmc68k	1	1.45 min	3.70 s	23.5×	1.000	pass
pbmc68k	32	1.45 min	3.71 s	23.4×	1.000	pass

Integration

1 task

Seurat2/2 pass

integrate_cca

Best8.32×

Median8.10×

Datasets1

ifnb · 32 threads1.82 min to 13.14 s

ifnb1, 32 threadsmin concordance 0.999 embedding_cor

Dataset	Threads	Baseline	Optimized	Speedup	Concordance	Pass
ifnb	32	1.82 min	13.14 s	8.32×	0.999	pass
ifnb	1	1.35 min	10.31 s	7.87×	0.999	pass

Raw benchmark runs98 of 98 shownShow table