Sandbox Provider Ratings

Each participant's average score for every benchmark in our stealth benchmarks over the last 7 days.

Composite Matrix

best per benchmarkLast updated November 5

Provider	Rating	Burst TTI	DAX	Snap & restore	OCI boot	OS caps	Burst Survival	CPU Burst	CPU	Fidelity	FS API	FS I/O	Pause / resume	Rootfs	Snap fanout	Snapshot	Soak	Volumes
1Vercel	81.3	96	73	98	—	88	100	57	37	98	63	94	98	86	97	98	31	91
2Superserve	79.1	93	80	100	—	86	100	67	4	100	100	97	86	83	100	100	1	—
3Novita	77.9	83	59	84	97	94	100	—	4	73	—	—	99	97	79	71	31	61
4Blaxel	77.8	95	70	82	—	86	100	51	8	70	100	97	0	50	90	79	100	99
5Daytona	75.8	93	76	91	95	67	100	45	39	95	51	96	0	88	96	96	100	54
6Freestyle	73.5	71	0	99	—	98	100	—	7	79	—	—	0	—	97	99	100	—
7Tensorlake	68.7	88	0	98	—	77	100	54	61	98	68	97	0	94	98	98	31	—
8E2B	67.8	81	53	99	99	76	100	63	2	62	63	0	99	85	99	99	32	—
9Modal	67.6	92	68	100	99	80	100	78	6	91	100	0	0	90	100	99	31	8
10Isorun	65.2	99	75	100	100	84	100	0	14	0	14	66	100	100	100	100	31	—
11GravixLayer	63.9	69	34	0	—	65	88	—	0	69	—	—	83	—	85	80	71	—
12Upstash	63.1	73	81	99	—	0	100	57	8	0	8	96	0	99	99	99	100	—
13Buddy Works	61.0	—	—	0	—	72	100	—	67	0	—	—	0	—	98	99	95	—
14Mosaic	60.2	91	81	0	—	73	99	48	13	0	8	97	0	94	89	96	100	—
15CreateOS	59.7	93	71	81	—	96	100	56	24	0	83	97	0	86	0	0	100	62
16Promigence	58.4	98	90	0	98	91	100	—	3	0	—	—	100	—	0	0	33	—
17Sail	55.4	41	77	0	—	98	87	72	8	0	61	96	97	100	0	0	100	—
18SFCompute	54.8	95	0	0	97	84	100	—	92	0	—	—	0	—	98	99	30	—
19Tenki	51.6	88	54	0	—	76	94	59	59	0	32	96	96	69	0	0	36	72
20Namespace	42.5	39	61	0	95	72	100	49	23	0	0	98	0	100	0	0	42	71
21Cloudflare	42.0	93	65	0	—	86	100	49	44	0	18	89	0	45	0	0	24	—
22Microsandbox	39.6	49	71	0	79	74	87	43	13	0	2	84	0	50	0	0	86	0
23Beam	38.7	89	69	0	99	83	100	81	6	0	8	3	0	82	0	0	31	0
24Runcloud	37.8	0	0	0	—	90	98	50	63	0	2	0	96	75	41	52	31	—
25Sprites	35.4	5	0	0	—	85	43	76	69	0	71	94	0	58	0	0	30	—
26Arker	33.8	96	56	0	—	0	100	46	12	0	2	96	0	94	0	0	32	—
27Sandbox0	32.5	3	45	0	—	83	100	50	30	0	32	62	0	45	0	0	31	—
28Miosa	30.9	99	76	0	—	70	100	0	26	0	0	0	0	98	0	0	30	—
29CodeSandbox	29.7	0	0	88	—	0	98	60	11	0	32	0	87	24	0	0	30	—
30Archil	25.5	69	0	0	96	0	75	47	2	0	5	69	0	37	0	0	30	0
31Railway	25.3	31	0	0	—	64	75	—	6	0	—	—	0	—	0	0	71	—
32Google Cloud Run	23.0	67	0	0	—	0	100	45	0	0	0	85	0	0	0	0	30	0

Methodology

This page is a meta-benchmark: it never starts a sandbox. It reads the daily composite scores from our stealth benchmarks and averages them into one weekly number per provider. Stealth benchmark subscribers can view the detailed results and use them to improve their infrastructure.

How the Weekly Composite Is Built

The window is the previous 7 complete days. Each provider's score is averaged in three steps:

Per benchmark, per day — average the composite score across all of that benchmark's runs that day, so re-runs are folded in rather than cherry-picked.
Per day — average those numbers across every sandbox benchmark that scored the provider that day.
Per week — average the daily composites across the days that had data. That is the weekly composite shown on each card and in the Composite score row.

Missing Data

Missing data is skipped, never zeroed. A benchmark with no runs on a given day doesn't count toward that day, and a day with no scores doesn't count toward the week.

Each matrix cell is that benchmark's average across the days it scored the provider; hover a cell to see how many days of the window it covers. A — means the provider has no score for that benchmark this week.