Prometheus v3.14.0 released on 18-08-2026
Prometheus v3.14.0 is out now. It tightens API semantics, advances PromQL (including duration expressions and new experimental functions), and brings TSDB and scraper hardening that reduce the risk of data-loss and restart corruption.
See the Prometheus GitHub release notes and the project documentation for full details and upgrade instructions.
What’s in this release
- API changes: the /api/v1/query and /api/v1/query_range stats parameter is deprecated for values other than true and all; /api/v1/status/config now shows explicitly-set empty separator and replacement in relabel configs.
- PromQL and UI: duration expressions are enabled by default (the promql-duration-expr flag is now a no-op), first_over_time is promoted to stable, and experimental features like start_timestamp(instant-vector) and optional start-timestamp extrapolation for rate()/increase() are available behind feature flags. The UI now supports duration expressions with syntax highlighting, autocompletion and linting for range selectors and subqueries.
- TSDB, histograms and remote write/read: experimental encoding of histogram start timestamps (behind histograms-st-encoding) and forwarding of histogram start timestamps in remote-write V2; new TSDB metrics expose native histogram series and bucket counts; multiple fixes to prevent silent data loss, restart-related corruption and resource leaks.
Upgrade notes
- Deprecation: update clients and tooling that set stats to values other than true or all. Other values still work but emit a deprecation warning and will be rejected in the next major release; check API callers and change them to true/all or remove the parameter.
- Feature flags and rollback: some capabilities are experimental and require feature flags (for example histograms-st-encoding and use-start-timestamps). If you encounter regressions, revert to the previous release while you adjust flags and client code.
Tell the community how the upgrade went or report any issues on the Prometheus issue tracker and discussion channels; user feedback helps catch edge cases that automated tests miss.
