$
npx @buildinternet/releases get dbt-labsRecently Shipped16 releases · updated Oct 14, 2026
Recently shipped native Amazon Redshift data sharing, plus a dbt-core 1.13 beta that bundles dbt State.
Redshift models can now read and write shared data. dbt materializes across Redshift clusters, serverless workgroups, and AWS accounts without copying data. dbt-redshift also moved off legacy Postgres metadata APIs onto Redshift-native SVV_* views and SHOW APIs, giving faster metadata resolution and better behavior under high concurrency.
The Summit launch set the v2 baseline. dbt v2 and dbt State went GA alongside Fivetran Context Layer, dbt Charts, new Wizard experiences, and an open lakehouse vision. dbt State runs on change rather than schedule, works in your own orchestrator (Airflow, Dagster, GitHub Actions, a laptop) and on Snowflake's dbt projects, across Snowflake, BigQuery, Databricks, and Redshift. v2 added AgentSkills-format SKILL.md installs gated on ai_provider, native Databricks metric views, and an adbc_clickhouse driver. Lake Compute reads Unity Catalog with Databricks PAT or OAuth M2M, read-only; existing local Unity attach options must move from config.lakecompute to config.duckdb.
dbt-core 1.13 beta turns on older flags and pre-bundles State. Breaking: flags introduced in 1.9 and 1.10 now default to true. The dbt-state plugin ships as a dependency but stays opt-in — enable it with --manage-state, DBT_ENGINE_MANAGE_STATE=true, or manage_state in dbt_project.yml. --use-v2-parser delegates parsing to the fusion parser, and MAXIMUM_SEED_SIZE_MIB is now configurable.
Maintenance. v2.0.8 lowered the minimum Python for the dbt and dbt-oss wheels to 3.10 and restored dbt build --write-catalog writing catalog.json. v2.0.5 fixed a BigQuery incremental path that could use REPLACE TABLE instead of MERGE.
A new datasharing: true profile credential switches the dbt-redshift adapter's metadata queries to Redshift-native SVV_* system views and SHOW APIs, making cross-database writes succeed and unlocking dbt Mesh cross-project refs, Advanced CI, and workload isolation for Redshift data sharing. The same migration away from legacy Postgres pg_* catalog views speeds up metadata resolution under high concurrency. The flag is currently in Beta on the Latest release track and not available on Compatible or Extended tracks.
dbt State GA: models reuse instead of rebuilding
BreakingThis release1 featureNew capabilitiesAI-tallied from the release notesdbt State is generally available, comparing each node's compiled SQL and upstream freshness to reuse, clone, or rebuild models at the node level, with pricing at $0.094 per daily active target table and each test counted separately. Views using select *, non-deterministic Jinja, and BigQuery external sources without loaded_at_field or loaded_at_query are always rebuilt; former state-aware orchestration users must map build_after to lag_tolerance and require_fresh_data_from, defaulting to 45m and any when build_after is absent.
Lake Compute reads Unity Catalog; module regex flags fixed
BreakingpreThis release1 featureNew capabilities2 enhancementsImprovements to existing features6 fixesBug fixesAI-tallied from the release notesdbt-core · v2.0.0-dev.41-test-pypi
Lake Compute can now read AWS Databricks Unity Catalog tables using Databricks profile PAT or OAuth M2M credentials, with declaration-aware validation and dbt debug attach parity. Fixed modules.re functions silently dropping keyword arguments like flags, maxsplit, and count, plus Databricks tag diffs that emitted unchanged columns and keys.
BigQuery incremental tables no longer REPLACE instead of MERGE
This release1 fixBug fixesAI-tallied from the release notesdbt-core · v2.0.5
Both Legacy and Foundry BigQuery drivers now continue on ADBC.GetObjects 404/403 responses, fixing a case where incremental tables could be written with REPLACE TABLE instead of MERGE.
dbt v2 and dbt State GA; Fivetran Context Layer, dbt Charts debut
This release6 featuresNew capabilitiesAI-tallied from the release notesdbt v2, a full Rust rewrite of the dbt engine that parses a 10,000-model project up to 10x faster than v1, and dbt State, which checks warehouse metadata and model SQL to build, skip, clone or defer each run, are both generally available. Also announced: Fivetran Context Layer (Private Beta) with the open source Agents Schema standard, dbt Charts (Public Beta) for governed BI as version-controlled YAML, dbt Wizard for the dbt platform (Public Preview) plus Wizard CLI (Public Beta) and Wizard Desktop (Private Beta), Lake Compute (Private Beta) on DuckDB against Apache Iceberg tables, and an open lakehouse vision for customer-owned storage and compute.
dbt State GA: runs on change, not schedule
This release3 featuresNew capabilitiesAI-tallied from the release notesdbt State is generally available and reads model SQL plus warehouse metadata each run to decide whether to build, skip, clone, or defer a node, without selection syntax or manifest scripts. Pricing is consumption-based on daily active target tables, with every reuse after the first in a day free and a 30-day trial. It supports Snowflake dbt Projects, includes Cost Insights, a richer dbt state explain experience, and concurrent build management.