The real difference is who writes the tests. Great Expectations gives you a Python vocabulary for writing expectations. DataOps TestGen profiles your tables and writes most of the tests for you. Pick based on how many tables you have and who on your team writes Python.
Side by side, as of October 2026
Capability ratings come from DataKitchen’s head-to-head TestGen vs GX OSS comparison, reviewed June 11, 2026. We sell one of these, so check the matrix against your own install.
| DataOps TestGen (open source) | GX Core | |
|---|---|---|
| How tests get created | Generated from a profile of 55 column characteristics | You write expectations in Python |
| Auto-generated tests | Yes | No |
| Data profiling | Yes | Yes |
| Anomaly detection | Yes | Rule-based |
| Freshness, volume, schema monitors | Yes | Partial |
| Data quality scoring | Yes | No |
| AI agent integration (MCP) | Yes, 96 tools | No |
| Interface | Application with a UI | Python library |
| License | Apache 2.0 | Apache 2.0 |
| Free tier limit | 1 user, 1 connection, 1 project | None on the software |
| Paid edition | Enterprise, $100 per user and per connection per month | None since GX Cloud shut down June 1, 2026 |
| Steward | DataKitchen | Fivetran, since May 2026 |
When GX Core is the better choice
GX Core wins in three situations, and we’d tell you so on a sales call.
Your team already has a working expectation suite in Git, wired into pipeline runs, and it catches what you need. Rewriting working tests to switch tools is cost with no new coverage.
Your checks belong in Python code next to the pipeline step they guard, and the people writing them are engineers who want them reviewed like any other code.
You have several engineers and no budget. GX Core costs $0 in software at any team size. Open-source TestGen stops at one user and one database. Two engineers on one database pay $300 a month for TestGen Enterprise.
When TestGen is the better choice
When the problem is coverage. If you have hundreds of tables and nobody has time to write expectations for them, a framework that speeds up writing doesn’t change the math. DataKitchen’s argument for generating 80% of tests is that Great Expectations, Soda Core, and dbt tests are all still code you write, one test at a time.
TestGen also suits teams where data stewards review results without writing Python, and teams that want monitors, scorecards, and a catalog without assembling them. It connects to an AI client over MCP, which GX Core does not.
Generated tests have a limit. Know it before you trust a green run. In DataKitchen’s NYC taxi exercise, 62 generated tests passed on a 300,000-row sample, because the minimum-fare test calibrated its floor to negative $367, the lowest fare present. On the full month of 2,964,624 rows, 11 of 69 tests failed, including 4 generated minimum-value tests, and all 7 business rules an AI wrote from the profile failed. Generation gets you coverage. Business rules still need someone who knows what correct means.
If you were on GX Cloud
That decision got made for you. GX Cloud shut down on June 1, 2026, and its customers had about 30 days to move. The GX Cloud comparison page covers the replacement question.
GX Core continuing under Fivetran is good news for the library. DataKitchen’s post on venture-funded open source adds a caution: Fivetran said it “intends to support ongoing project maintenance,” which is a statement of intent from the company that now owns the project. TestGen comes from a company that has been profitable since 2013 with no outside investment. Weigh the steward along with the features.