Transparency alternatives to Union
Other MLOps and model serving vendors we measure, sorted by how much we found at them and not at Union. This is a comparison of disclosure, not of features or fitness.
12 shown of 34 that differ from Union on at least one criterion we measured for both. Public transparency score, not a product or security rating.
Union has 7 gaps in what we could measure: Data processing agreement, Subprocessor list, Status page with history, No third-party tracking before consent, security.txt (RFC 9116), Notice before subprocessors change, Named privacy contact. The column “Found there, not found at Union” is the one that closes them.
| Vendor | Score | Found there, not found at Union | Found at Union, not found there | |
|---|---|---|---|---|
| Superwisesuperwise.aiSee also Transparency alternatives to Superwise | 65/100 | 4 Subprocessor list, security.txt (RFC 9116), Notice before subprocessors change, +1 | 1 Uptime SLA with a figure | Compare |
| H2O.aih2o.aiSee also Transparency alternatives to H2O.ai | 57/100 | 4 Data processing agreement, security.txt (RFC 9116), Notice before subprocessors change, +1 | 2 Public pricing, Uptime SLA with a figure | Compare |
| Basetenbaseten.coSee also Transparency alternatives to Baseten | 52/100 | 4 Data processing agreement, Status page with history, Notice before subprocessors change, +1 | 3 Data location stated, Certifications named, Uptime SLA with a figure | Compare |
| RunPodrunpod.ioSee also Transparency alternatives to RunPod | 52/100 | 4 Data processing agreement, security.txt (RFC 9116), Notice before subprocessors change, +1 | 2 Data location stated, Uptime SLA with a figure | Compare |
| Nscalenscale.comSee also Transparency alternatives to Nscale | 52/100 | 3 Subprocessor list, Status page with history, security.txt (RFC 9116) | 2 Public pricing, Uptime SLA with a figure | Compare |
| Nebiusnebius.comSee also Transparency alternatives to Nebius | 30/100 | 3 Status page with history, security.txt (RFC 9116), Named privacy contact | 3 Data location stated, Public pricing, Uptime SLA with a figure | Compare |
| Vast.aivast.aiSee also Transparency alternatives to Vast.ai | 52/100 | 2 Data processing agreement, Named privacy contact | 1 Uptime SLA with a figure | Compare |
| Modalmodal.comSee also Transparency alternatives to Modal | 39/100 | 2 Data processing agreement, Notice before subprocessors change | 3 Data location stated, Certifications named, Uptime SLA with a figure | Compare |
| DataRobotdatarobot.comSee also Transparency alternatives to DataRobot | 30/100 | 2 Status page with history, Named privacy contact | 3 Data location stated, Certifications named, Uptime SLA with a figure | Compare |
| Replicatereplicate.comSee also Transparency alternatives to Replicate | 30/100 | 2 Status page with history, Named privacy contact | 3 Data location stated, Certifications named, Uptime SLA with a figure | Compare |
| Dataikudataiku.comSee also Transparency alternatives to Dataiku | 26/100 | 2 Data processing agreement, security.txt (RFC 9116) | 4 Data location stated, Public pricing, Certifications named, +1 | Compare |
| Lambdalambdalabs.comSee also Transparency alternatives to Lambda | 17/100 | 2 security.txt (RFC 9116), Named privacy contact | 4 Data location stated, Public pricing, Certifications named, +1 | Compare |
Only criteria we could measure for both are counted. Where we could measure one and not the other, the row is silent: that would be a statement about our measurement, not about either vendor. Every single result with its source is on the two profiles.