Data Engineering · 2 min read

dbt v2 reaches GA with Rust engine and 30–40% faster end-to-end runs

By Tom Whitfield413 words

What's new in dbt: November 2026
What's new in dbt: November 2026 — geralt / pixabay

By Tom Whitfield

dbt v2 is generally available (GA) on five data platforms, with more than 14,000 active projects running it weekly. The Nov. 17, 2026 update expands local testing, dbt State and AI tooling. dbt claims 30–40% faster end-to-end runs for customers but discloses no customer mix or benchmark method.

The Nov. 17 update came two months after dbt Summit, attended by thousands in Las Vegas and online. dbt framed a shift from analytics engineers to context engineers, and from the modern data stack to open data infrastructure. Agents write more SQL while users increasingly query data through AI.

What changed in dbt v2?

Compared with v1, dbt v2 uses Rust (a systems programming language), scales to thousands of models, adds SQL comprehension (automatic interpretation of SQL) and column-level lineage, and emits queryable project metadata for agents and governance tools at compile time. dbt says self-hosted use remains open source forever.

GA covers Amazon Redshift, Databricks, Google BigQuery, Snowflake and DuckDB; ClickHouse and Apache Spark remain in Beta. Matt Karan, senior data engineer at Obie, said: “Everything about v2 has sped up our workflows.” That testimony cannot establish typical savings.

How do local and batch tests reduce warehouse work?

Local unit tests are in Alpha (early testing). dbt v2 translates SQL to DuckDB and runs tests wherever dbt runs, avoiding a warehouse round trip and the latency and compute charges it adds to every iteration. Teams set DBT_ENGINE_EXPERIMENTAL_LOCAL_UNIT_TESTS=true, then add compute: local to one test or +compute: local project-wide.

Batch testing addresses projects that can generate hundreds of small warehouse queries per run. Now GA, it combines each model’s unique and not_null tests into one query per test type without changing test definitions.

dbt’s internal tests cut runtime 40% and warehouse queries 65%. Enable --batch-tests with dbt build or dbt test, or set batch_tests: true under flags in dbt_project.yml. The company reports no external benchmark.

What does dbt State change?

dbt State is GA and rebuilds only changed models. Lag tolerance recommendations, also GA, tune sub-project and model-level freshness SLAs (service-level agreements), preview lag_tolerance effects and can skip more work within tolerance limits.

The former 30-day trial now rolls into a free plan with monthly allotments measured in daily active target tables (DATT):

  • 1,000 DATT for self-hosted dbt.
  • 1,000 DATT on dbt platform Developer and Starter.
  • 2,000 DATT on Enterprise and Enterprise+.

docs.getdbt.com (Original)

Price tags

Share this article:

More from Tom Whitfield

Tom Whitfield

Show full bio

Staff writer covering regulation, policy and their impact on Analytics Digest is a professional daily at Analytics Digest.

16 articles

Read also

« Previous articleNext article »