Airbyte vs Fivetran

Quick verdict

Airbyte is a data integration (ELT) platform. Its free, open source edition, Airbyte Core, runs on Kubernetes in your own infrastructure; Airbyte documents a local install with its abctl tool and Docker, and recommends 4 or more CPUs and 8 GB of memory. The managed offerings are the cloud plans Standard, Plus and Pro, plus Enterprise Flex, a hybrid option with separate data planes for compliance and data sovereignty. According to Airbyte's licences page, the platform and the connectors in its public repositories are licensed under the Elastic License 2.0 (ELv2), and only the Airbyte Protocol is under MIT; community-built connectors may carry their own licences. The latest Self-Managed release listed in Airbyte's release notes is 2.3 (15 September 2026).

Fivetran is a fully managed data movement service: Fivetran runs and maintains the connectors, and you configure sources and destinations. It states 750+ sources and 200+ destinations, sells Free, Standard, Enterprise and Business Critical plans, and bills on Monthly Active Rows (MAR). In 2025 it signed an agreement to acquire Census (reverse ETL, announced 1 May 2025) and acquired Tobiko Data, the company behind SQLMesh and SQLGlot (announced 3 September 2025). On 1 June 2026 it completed an all-stock merger with dbt Labs; the combined company initially operates as Fivetran + dbt Labs, with George Fraser as CEO and Tristan Handy as President.

Side by side

Key differences

Open source and self-managed versus fully managed

The deepest difference is who runs the software. With Airbyte Core you install the platform on your own Kubernetes cluster (or locally through abctl), and you are responsible for upgrades, scaling, monitoring and security. In return, data never has to leave your infrastructure, there is no per-row charge, and you can read and change the code. The platform licence is ELv2: Airbyte's licence FAQ says you may use it commercially and build products on it, but you may not offer Airbyte itself to others as a managed service, circumvent licence-key features, or remove licence notices. Airbyte's licences page places its connectors under ELv2 as well; only the Airbyte Protocol is MIT.

Fivetran is SaaS only. Its documentation describes two deployment models: SaaS, where all processing happens in the Fivetran cloud on a cloud provider you choose, and Hybrid Deployment, where data is processed inside your own network while Fivetran's cloud orchestrates and configures the pipelines. The pricing page lists Hybrid Deployment from the Enterprise plan. In our view, Fivetran suits teams that would rather pay than run pipeline infrastructure; Airbyte suits teams that have platform engineers or strict data residency needs and want the self-hosted path available.

Connector catalogues and who maintains them

Raw connector counts are close (Airbyte states 700+, Fivetran 750+ sources), so the more useful question is who maintains each connector. Airbyte documents several support levels. Airbyte (certified) connectors are maintained by Airbyte, described as production ready, and covered by its support. Enterprise connectors are built by Airbyte for critical enterprise systems and are available only to Pro and Enterprise customers, at additional cost. Marketplace connectors are maintained by community members, are not maintained by Airbyte, are not covered by support SLAs, and may change without notice. Custom connectors, built with the Connector Builder or CDK, are maintained by you. Low-usage connectors can be archived.

Fivetran groups its catalogue into Standard connectors (managed connectors for common SaaS, ERP, database, event and file sources), Lite connectors (built for specific use cases with niche applications), partner-built connectors (maintained by technology partners) and the Connector SDK for code you write yourself. Enterprise database connectors are listed from the Enterprise plan. So both catalogues have a vendor-maintained core and a long tail maintained by someone else; check the support level of every connector you depend on, not the headline count.

Change data capture for databases

Airbyte documents CDC support for Postgres, MySQL, Microsoft SQL Server, MongoDB, Oracle DB, SAP HANA and IBM Db2 (check each connector's support level; some enterprise sources are Enterprise connectors). Its CDC documentation lists general limitations: CDC incremental generally needs a primary key, works on tables not views, captures only INSERT, UPDATE and DELETE, starts with an initial snapshot, and a final table will not show rows whose latest change was a delete. For SQL Server, CDC must be enabled on the source database first, for example:

-- T-SQL (SQL Server): enable CDC on a database and one table
EXEC sys.sp_cdc_enable_db;
EXEC sys.sp_cdc_enable_table
     @source_schema = N'dbo',
     @source_name   = N'orders',
     @role_name     = NULL;

Fivetran documents that its database connectors use each database's native change mechanism (binary log, Change Data Capture, change streams, logical replication and others, depending on the engine), and offers Teleport Sync, a proprietary incremental method that needs only a read-only SQL connection. Fivetran says all database connectors sync deleted data, marking deleted rows as deleted in the destination rather than removing them. Fivetran also states that initial (historical) syncs are free of MAR charges. Both tools therefore cover the common engines; the practical differences are deletion handling, which plan includes the connector you need, and whether you want to run the CDC pipeline yourself.

Pricing models: credits and capacity versus monthly active rows

Airbyte Cloud has two models. Standard and Plus are billed by data volume through credits (Standard has a monthly base with extra credits; Plus sells fixed monthly credit packages). Pro and Enterprise Flex are capacity based, priced by Data Workers, which Airbyte describes as dedicated compute units, each typically running about three syncs at once, with no charges based on data volume. Self-managed Core has no licence fee; you pay for the infrastructure you run it on.

Fivetran bills on Monthly Active Rows: the number of distinct rows (by primary key) inserted or updated in the destination in a calendar month, each counted once a month however often it changes. Since 2025 MAR is counted per connection, so two connections syncing the same data both count. Paid plans add a base charge for each standard connection with between 1 and 1 million MAR in a month, and dbt transformations are billed by model runs above a free monthly allowance. In our view, MAR pricing rewards stable data and penalises high-churn tables, while Airbyte's capacity model trades volume sensitivity for paying for compute whether or not it is busy. Exact prices are in the pricing section below.

Transformations and dbt

Fivetran runs transformations as part of the product: Quickstart data models (no-code models for supported connectors), Fivetran data models for dbt projects, and Transformations for dbt hosted by Fivetran, scheduled after syncs, on a frequency, or from the project's deployment.yml. It also integrates with dbt Cloud and Coalesce. Since the merger completed in June 2026, dbt is part of the same company; the merger announcement released dbt Core v2.0 (alpha), which open sources the dbt Fusion engine runtime under Apache 2.0, and says the combined company works across any cloud, engine and tool.

Airbyte has moved away from running transformations itself. Its April 2024 release notes state that, as of v0.57.0, custom dbt transformations were deprecated in Open Source, and its current product documentation does not describe a built-in transformation step. Teams typically run dbt (Core or Cloud) after Airbyte syncs, triggered by an orchestrator such as Apache Airflow, Dagster or Prefect; see Apache Airflow vs Airbyte. dbt itself works with data loaded by either tool. A typical first model on raw loaded data looks like this:

-- dbt model (Snowflake SQL): deduplicate raw orders loaded by an ELT tool
select *
from {{ source('raw', 'orders') }}
qualify row_number() over (partition by order_id order by updated_at desc) = 1

Pricing and licensing

Airbyte. Listed on airbyte.com/pricing in October 2026 (USD): Standard starts at USD 20 per month, with extra credits at USD 5 each, billed by data volume; Plus is sold as monthly credit packages from USD 189 (40 credits) to USD 4,999 (2,000 credits); Pro and Enterprise Flex are capacity based (Data Workers) with custom quotes. Standard and Plus come with a 14-day free trial including 400 credits. Airbyte Core is free to self-host. The page did not state a fixed volume per credit, so use Airbyte's calculator for your own sources.

Fivetran. Listed on fivetran.com/pricing in October 2026: plans are Free, Standard, Enterprise and Business Critical, with usage priced on MAR and discounts for annual contracts. The Free plan includes 500,000 MAR for connections, 3,500 MAR for activations and 5,000 monthly model runs, with no base charge. On paid plans, a USD 5 base charge applies to each standard connection with between 1 and 1 million MAR in a month. As one example from Fivetran's own pricing page, a Marketo connection with 847,574 MAR is shown at USD 424.54 for the month. Transformation model runs are free up to 5,000 a month, then tiered from USD 0.01 per run down to USD 0.002 per run above 100,000. Initial syncs and new connections' 14-day trial periods do not incur MAR charges. Per-MAR rates depend on plan and volume; use Fivetran's estimator rather than extrapolating from the example.

Pricing checked on the vendors' official pages on 7 October 2026. Prices change; confirm before buying.

Where each one leads

Limitations

When to choose each

When neither is right

• You mainly need to schedule and orchestrate jobs, not ingest data: see Apache Airflow vs Airbyte and Apache Airflow vs Fivetran.
• You need heavy in-pipeline transformation with a visual designer: compare Fivetran vs Matillion, Airbyte vs Matillion and Airbyte vs Talend.
• You want a lightweight, code-first open source alternative built on Singer taps: see Airbyte vs Meltano.
• You only move data between two databases on the same platform: native replication or a scheduled INSERT ... SELECT may be enough, without an ELT service.

Final recommendation

Frequently asked questions

Sources

• Airbyte pricing
• Airbyte documentation: platform overview
• Airbyte: licence FAQ
• Airbyte: connector support levels
• Airbyte: change data capture
• Airbyte: Self-Managed release notes
• Airbyte: April 2024 release notes (dbt transformations deprecation)
• Fivetran pricing
• Fivetran docs: usage-based pricing (MAR)
• Fivetran docs: database connectors
• Fivetran docs: deployment models
• Fivetran docs: transformations
• Fivetran press release: Fivetran and dbt Labs complete merger (1 June 2026)
• Fivetran press release: agreement to acquire Census (May 2025)
• Fivetran press release: acquisition of Tobiko Data (September 2025)

Checked October 2026.

How we research comparisons: our editorial method.