ClickHouse has expanded its open-source JSON database benchmark from five systems to 27, and says it still leads the billion-document comparison on identical hardware. The useful change is a broader, reproducible comparison of how database engines handle large volumes of real-world JSON, with ClickHouse’s own performance claims clearly in view.
ClickHouse Watch analysis
What happened
ClickHouse says the updated JSONBench tests the original five systems again using newer versions and updated configurations, and adds 22 more database engines. Its current comparison includes 14 selected systems; the company says ClickHouse leads that field too. The benchmark uses real Bluesky events with their original nested JSON structure. ClickHouse says its query returned results in 661 milliseconds on a cold run, compared with 32 minutes and 26 seconds for MongoDB on the same machine type and published configurations.
The benchmark is open source, with its data, queries, configurations and results available to reproduce or inspect. ClickHouse also says it has shipped 178 changes to its native JSON type since the first comparison, including faster reads, lower memory use, new indexes and direct JSON support in Parquet. Read ClickHouse’s JSONBench update.
Why it matters
JSON is common in logs, events and applications, but flexible documents can make analytical work awkward. This comparison gives developers a concrete starting point for evaluating systems against the same workload, rather than relying on feature lists or a benchmark with an opaque recipe.
The striking time comparison is ClickHouse’s result from its own benchmark, not an independent verdict on every database or workload. The wider field and public methodology make the claim easier to test, which is more useful than asking readers to take a chart on faith.
Our read
A benchmark that can be rerun and challenged is worth more than a victory lap with no working. ClickHouse has made the comparison broader and put its recipe in public; now the community has a fair invitation to reproduce it, improve configurations and add missing systems. Treat the headline gap as a result to investigate, not a universal promise about your own data.
What to watch
- Whether contributors reproduce the results or identify improvements to the test configurations.
- How the other engines perform across the full set of queries, loading and storage measures.
- Whether future ClickHouse releases continue to improve JSON performance without compromising flexibility.
Discussion spark: When choosing a database for JSON analytics, should a reproducible vendor-run benchmark carry real weight, or should buyers wait for independent tests?
Sources and evidence
- Source update (16 November 2026, 09:25 UTC)
Independent WittyWires tracker for public updates about ClickHouse. Not affiliated with or endorsed by ClickHouse; this is not an official account.