/

Airbyte

Filament vs. Airbyte

Side-by-side capabilities and benchmarks.

Introduction

Airbyte is a data integration platform. It is known for a large catalog of connectors, most of them built and maintained by its community. It is source-available and runs on Kubernetes or as a hosted service. Filament is an open-source replication engine that moves data into the systems you already run, verifies every batch it writes, and resumes where it left off when a run fails.

HIGHLIGHTS

90.5× faster

ROWS PER SECOND

Filament

2,597,498 ROWS

Airbyte

28,698 ROWS

NYC Taxi, Postgres -> Postgres

90.5× faster

ROWS PER SECOND

Filament

2,597,498 ROWS

Airbyte

28,698 ROWS

NYC Taxi, Postgres -> Postgres

32.3× FASTER

WALL TIME

Filament

3.79 SEC

Airbyte

122.4 SEC

TPC-H, Postgres -> Postgres

32.3× FASTER

WALL TIME

Filament

3.79 SEC

Airbyte

122.4 SEC

TPC-H, Postgres -> Postgres

36.3× less

CPU SECONDS

Filament

377.4 SEC

Airbyte

13703.3 SEC

NYC Taxi, Postgres -> Postgres

36.3× less

CPU SECONDS

Filament

377.4 SEC

Airbyte

13703.3 SEC

NYC Taxi, Postgres -> Postgres

4.3× less

PEAK MEMORY

Filament

2.97 GIB

Airbyte

12.82 GIB

NYC Taxi, Postgres -> Postgres

4.3× less

PEAK MEMORY

Filament

2.97 GIB

Airbyte

12.82 GIB

NYC Taxi, Postgres -> Postgres

Galaxy product illustration

License and control

License

Apache 2.0, open source.

Elastic License 2.0, source-available. Connectors moved from MIT to ELv2 in July 2025.

License

Apache 2.0, open source.

Elastic License 2.0, source-available. Connectors moved from MIT to ELv2 in July 2025.

Self-hosted in full.

Yes. One repository, no paid tier, nothing gated.

Core edition only. Single user, one workspace, no RBAC, no SSO, Slack-only support. Self-Managed Enterprise is no longer sold.

Self-hosted in full.

Yes. One repository, no paid tier, nothing gated.

Core edition only. Single user, one workspace, no RBAC, no SSO, Slack-only support. Self-Managed Enterprise is no longer sold.

Adding a new source

A YAML manifest for REST APIs, loaded at runtime with no rebuild. A Go driver for databases and other driver-based sources.

Connector Builder UI or low-code YAML, packaged as its own Docker image. Python CDK or Kotlin Bulk CDK for the rest.

Adding a new source

A YAML manifest for REST APIs, loaded at runtime with no rebuild. A Go driver for databases and other driver-based sources.

Connector Builder UI or low-code YAML, packaged as its own Docker image. Python CDK or Kotlin Bulk CDK for the rest.

Connector maintenance

Every connector lives in the engine repository with a published maturity level (alpha, beta, stable).

59 certified sources maintained by Airbyte. 535 marketplace sources maintained by the community with no SLA. 6 enterprise sources, maintained by Airbyte, cost extra.

Connector maintenance

Every connector lives in the engine repository with a published maturity level (alpha, beta, stable).

59 certified sources maintained by Airbyte. 535 marketplace sources maintained by the community with no SLA. 6 enterprise sources, maintained by Airbyte, cost extra.

Services to operate

Server and control plane, with Postgres and NATS. Each run is its own Kubernetes Job. One binary for evaluation.

10 services enabled by default in the Helm chart (server, worker, workload launcher, workload API, Temporal, cron, manifest server, bootloader, Postgres, MinIO), plus a pod per sync.

Services to operate

Server and control plane, with Postgres and NATS. Each run is its own Kubernetes Job. One binary for evaluation.

10 services enabled by default in the Helm chart (server, worker, workload launcher, workload API, Temporal, cron, manifest server, bootloader, Postgres, MinIO), plus a pod per sync.

Runtime

Static Go binaries on distroless images. One runner code path wherever a run executes.

Java 21 and Kotlin platform. Python or Java connectors. Every sync launches a pod with separate source, destination, and orchestrator containers.

Runtime

Static Go binaries on distroless images. One runner code path wherever a run executes.

Java 21 and Kotlin platform. Python or Java connectors. Every sync launches a pod with separate source, destination, and orchestrator containers.

Scheduling

Built-in cron per pipeline. Safe to run multiple control planes.

Built-in on Temporal. Schedule accuracy documented as plus or minus 30 minutes.

Scheduling

Built-in cron per pipeline. Safe to run multiple control planes.

Built-in on Temporal. Schedule accuracy documented as plus or minus 30 minutes.

SaaS source execution

Manifests interpreted inside the engine, in the same process as the run. No per-connector containers.

One container per source per sync. YAML connectors run inside a Python declarative runtime image.

SaaS source execution

Manifests interpreted inside the engine, in the same process as the run. No per-connector containers.

One container per source per sync. YAML connectors run inside a Python declarative runtime image.

Embeddable

Yes. Go library (app.Run) with swappable event bus, state store, and secrets.

PyAirbyte runs connectors from Python, with no scheduling, orchestration, or alerting.

Embeddable

Yes. Go library (app.Run) with swappable event bus, state store, and secrets.

PyAirbyte runs connectors from Python, with no scheduling, orchestration, or alerting.

Interfaces

Web UI, ConnectRPC API, CLI.

Web UI, REST API, Terraform provider. CLI is for installation only.

Interfaces

Web UI, ConnectRPC API, CLI.

Web UI, REST API, Terraform provider. CLI is for installation only.

Correctness

Resume after failure

Mid-table, from per-shard checkpoints, for upsert, incremental, and CDC runs. A failed run lands in a partial state and continues on re-request.

Resumable full refresh from an artificial cursor on Postgres, MySQL, MSSQL, Mongo, and Python CDK sources, when the destination supports refreshes. Other sources restart the stream.

Resume after failure

Mid-table, from per-shard checkpoints, for upsert, incremental, and CDC runs. A failed run lands in a partial state and continues on re-request.

Resumable full refresh from an artificial cursor on Postgres, MySQL, MSSQL, Mongo, and Python CDK sources, when the destination supports refreshes. Other sources restart the stream.

Per-batch integrity check

CRC32-C computed independently on the read side and the write side of every batch. A mismatch fails the run.

Record counts compared per state message. A mismatch is logged, not failed, unless a feature flag is switched on. No checksum.

Per-batch integrity check

CRC32-C computed independently on the read side and the write side of every batch. A mismatch fails the run.

Record counts compared per state message. A mismatch is logged, not failed, unless a feature flag is switched on. No checksum.

Change data capture

Native. Postgres logical replication (pgoutput) with a consistent snapshot bootstrap. MySQL binlog with GTID cursors.

Embedded Debezium for Postgres, MySQL, MSSQL, MongoDB. Oracle, SAP HANA, and Db2 CDC are enterprise connectors.

Change data capture

Native. Postgres logical replication (pgoutput) with a consistent snapshot bootstrap. MySQL binlog with GTID cursors.

Embedded Debezium for Postgres, MySQL, MSSQL, MongoDB. Oracle, SAP HANA, and Db2 CDC are enterprise connectors.

Schema drift

New columns added to the destination automatically. Add-only, nothing dropped, nothing rewritten.

New columns propagate. A type change writes row errors until you refresh the schema and refresh the stream.

Schema drift

New columns added to the destination automatically. Add-only, nothing dropped, nothing rewritten.

New columns propagate. A type change writes row errors until you refresh the schema and refresh the stream.

Parallel reads within a table

Yes. Large tables split into up to 64 shards (keyset, bitmap, or ctid) read concurrently.

Socket mode on Bulk CDK database sources, concurrency defaulting to 1. Socket mode is enabled today only for the MySQL source.

Parallel reads within a table

Yes. Large tables split into up to 64 shards (keyset, bitmap, or ctid) read concurrently.

Socket mode on Bulk CDK database sources, concurrency defaulting to 1. Socket mode is enabled today only for the MySQL source.

Row lineage

Every replicated row carries the run id and start time. CDC rows also carry operation, source position, and sequence.

_airbyte_raw_id, _airbyte_extracted_at, and _airbyte_meta per row.

Row lineage

Every replicated row carries the run id and start time. CDC rows also carry operation, source position, and sequence.

_airbyte_raw_id, _airbyte_extracted_at, and _airbyte_meta per row.

Cost

What drives billing

Nothing. Compute you already run.

Cloud: credits at $2.50 each. APIs and custom sources bill $15 per million rows. Databases and files bill $10 per GB of serialized records. Pro and Enterprise Flex bill on Data Worker capacity.

What drives billing

Nothing. Compute you already run.

Cloud: credits at $2.50 each. APIs and custom sources bill $15 per million rows. Databases and files bill $10 per GB of serialized records. Pro and Enterprise Flex bill on Data Worker capacity.

Paid tiers

None.

RBAC, SSO, and multiple workspaces are Pro or Enterprise Flex. 15-minute syncs start at Cloud Plus. Typing and deduplication add warehouse compute.

Paid tiers

None.

RBAC, SSO, and multiple workspaces are Pro or Enterprise Flex. 15-minute syncs start at Cloud Plus. Typing and deduplication add warehouse compute.

Comparison

Benchmarks

Benchmarks

benchmark

Wall time

Lower is better

Lower is better

Filament

Airbyte

NYC Taxi, PG → PG (298.3M rows)

114.8 s

10,393.3 s

TPC-H, PG → PG (8.7M rows)

3.79 s

122.4 s

Rows per second

Higher is better

Higher is better

Filament

Filament

Airbyte

Airbyte

Better

NYC Taxi, PG → PG

NYC Taxi, PG → PG

2,597,498

2,597,498

2,597,498

28,698

28,698

28,698

Filament 90.5×

TPC-H, PG → PG

TPC-H, PG → PG

2,288,144

2,288,144

2,288,144

70,783

70,783

70,783

Filament 32.3×

CPU seconds

Lower is better

Lower is better

Filament

Filament

Airbyte

Airbyte

Better

NYC Taxi, PG → PG

377.4

377.4

377.4

13,703.3

13,703.3

13,703.3

Filament 36.3×

TPC-H, PG → PG

17.9

17.9

17.9

229.0

229.0

229.0

Filament 12.8×

Peak memory

Lower is better

Lower is better

Filament

Airbyte

NYC Taxi, PG → PG (298.3M rows)

2.97 GiB

12.82 GiB

TPC-H, PG → PG (8.7M rows)

1.83 GiB

7.42 GiB

Last updated on August 31, 2026

Full Filament vs. Airbyte benchmark report

Full Filament vs. Airbyte benchmark report

Full Filament vs. Airbyte benchmark report

Migrate to

Migrate to

Copyright © 2026 Galaxy. All rights reserved.