Trino vs Presto: History, Differences, Which to Use (2026)

Quick verdict

Short answer

For most new deployments, Trino is the more common starting point: it releases frequently, documents a wide set of connectors and fault-tolerant execution, and has managed and commercial options from Starburst, while Amazon Athena engine version 3 incorporates Trino functionality. Choose Presto if you want the Linux Foundation-governed project, if you are interested in its C++ native workers (Presto C++, built on Velox), or if your platform already standardises on it, for example IBM watsonx.data. The two are no longer drop-in replacements for each other, so pick one and test your queries and connectors on it.

How we know: This comparison is research-based: history, governance, releases, connectors, Presto C++ status and managed offerings were checked against trino.io, prestodb.io materials in the Presto GitHub repository, both projects' GitHub release pages, and AWS, Starburst and IBM documentation in October 2026. We have not run benchmarks, and vendor performance claims are not repeated.

Presto began at Facebook in 2012, when Martin Traverso, Dain Sundstrom and David Phillips created it for interactive analytics over Facebook's Hadoop data warehouse. Today the project lives at prestodb.io and on GitHub as prestodb/presto, under the Apache License 2.0, governed by the Presto Foundation, which Facebook established at the Linux Foundation in September 2019. Presto still uses 0.x version numbers; the latest release on GitHub is 0.299 (August 2026).

Trino is the fork led by Presto's original creators. It was first called PrestoSQL, and was renamed Trino on 27 December 2020; the announcement on trino.io says the rename followed Facebook's trademark application for Presto and its enforcement through the Presto Foundation. Trino is Apache 2.0 licensed and supported by the non-profit Trino Software Foundation. It uses whole-number releases; the latest is Trino 483 (July 2026). Starburst, founded by Trino's creators, sells commercial products built on Trino.

Both are query engines, not databases: they store no data of their own and run SQL over catalogs that point at object storage, table formats such as Apache Iceberg and Delta Lake, and other databases. For how this differs from a warehouse that stores data, see Data Warehouse vs Database.

Side by side

Aspect	Trino	Presto
Origin	Fork by Presto's creators; PrestoSQL until December 2020	Created at Facebook in 2012; the continuing prestodb project
Governance	Trino Software Foundation (non-profit)	Presto Foundation at the Linux Foundation (since September 2019)
Licence	Apache License 2.0	Apache License 2.0
Latest release (October 2026)	483 (July 2026)	0.299 (August 2026)
Execution	Java coordinator and workers	Java coordinator; Java workers or Presto C++ native workers built on Velox
Table format connectors	Iceberg, Delta Lake, Hudi, Hive, plus a Lakehouse connector	Iceberg, Delta Lake, Hudi, Hive (Presto C++ documents Iceberg as read only)
Fault tolerance	Documented fault-tolerant execution with QUERY and TASK retry policies	Not compared here; check the Presto documentation for your version
Managed and commercial options	Starburst Galaxy (SaaS) and Starburst Enterprise; Trino on Amazon EMR; Athena engine version 3 incorporates Trino functionality	IBM watsonx.data (Presto Java and Presto C++ engines); Presto on Amazon EMR
Main trade-off	Java only; frequent releases mean regular upgrade work	Smaller connector list and commercial ecosystem; C++ workers still reaching parity

Key differences

History and governance: why there are two projects

The split is about people and naming rather than technology. Presto's creators left Facebook and, in January 2019, set up the Presto Software Foundation to run their fork, PrestoSQL. Facebook then established the Presto Foundation at the Linux Foundation in September 2019 and, according to Trino's announcement, applied for and began enforcing a trademark on the Presto name. The fork was renamed Trino in December 2020, and its foundation is now the Trino Software Foundation.

Since then the codebases have diverged. They share architecture (a coordinator that plans queries, workers that execute them, connectors that expose catalogs) and much of their SQL dialect, but features, configuration properties and connector behaviour differ. Amazon EMR, which offers both, notes that they cannot run on the same cluster at the same time and recommends Trino going forward.

Presto C++ (Prestissimo) and Velox

Presto's most distinctive work is Presto C++, developed under the name Prestissimo: a drop-in replacement for Presto's Java workers, written in C++ on the open source Velox execution library. It implements the same REST endpoints, so a Java coordinator schedules work on C++ workers, and no JVM is needed on worker nodes. The Presto documentation describes it as in active development and documents limits: Hive is supported for reads and writes, Iceberg tables for reads only, and JMX metrics come from the Java coordinator rather than the C++ workers. IBM watsonx.data offers a Presto (C++) engine alongside Presto (Java).

Trino remains a Java engine. In our view, Presto C++ is the main technical reason to look at Presto for a new deployment, but check its connector and feature coverage against your workload before relying on it, because parity with the Java workers is still a stated goal rather than a finished state.

Connectors, table formats and reliability features

Both engines query data in place through connectors, and both document connectors for Apache Iceberg, Delta Lake, Apache Hudi and Hive, plus relational databases such as MySQL, PostgreSQL, SQL Server, Oracle and Redshift, and systems such as Kafka, MongoDB, Cassandra, Elasticsearch, Pinot and Druid. Trino's list (Trino 483) also includes ClickHouse, DuckDB, Snowflake, BigQuery, MariaDB, SingleStore, OpenSearch, Loki, Exasol and a Lakehouse connector. Presto's documentation also covers BigQuery, ClickHouse, SingleStore, Kudu, SAP HANA, ScyllaDB, an Arrow Flight connector and Lance.

Trino documents fault-tolerant execution: a QUERY retry policy that retries whole queries, and a TASK policy that retries individual tasks using an exchange manager that spools intermediate data to S3, Azure, GCS, HDFS or local storage. It is supported with connectors including Hive, Iceberg, Delta Lake, BigQuery and the main relational databases, and is aimed at long-running batch queries.

-- Trino or Presto: one query across two catalogs
SELECT c.region, sum(o.total) AS revenue
FROM postgresql.public.customers AS c
JOIN iceberg.sales.orders AS o
  ON o.customer_id = c.id
GROUP BY c.region;

The catalog.schema.table naming is the same in both engines; each catalog is configured by a properties file that names a connector. Connector-specific SQL, such as table properties and time-travel syntax on Iceberg tables, can differ, so check each engine's connector documentation.

Managed services and commercial support

Trino has the larger commercial ecosystem. Starburst, founded by Trino's creators, sells Starburst Galaxy, a managed SaaS on AWS, GCP and Azure, and Starburst Enterprise for self-managed private cloud, hybrid and on-premises use. Amazon EMR ships Trino (EMR 7.14.0 includes Trino 479). Amazon Athena engine version 3 is AWS's own engine: AWS says it integrates improvements from the open source Trino and Presto projects, lists functionality incorporated from Trino and points to the Trino documentation for functions, but it does not support Trino or Presto connectors (Athena uses its own Federated Query instead) or Trino's fault-tolerant execution.

Presto is offered in IBM watsonx.data, which provides both Presto (Java) and Presto (C++) engines, and on Amazon EMR (EMR 7.14.0 includes Presto 0.287), although AWS recommends Trino for new EMR work.

Pricing and licensing

Trino and Presto are both free and open source under the Apache License 2.0. Running either yourself costs the compute and storage it uses, plus the effort of operating a cluster.

Managed options are priced by each provider. Starburst's site offered a Starburst Galaxy free trial with up to USD 500 in credits and no credit card in October 2026; Starburst Enterprise is sold by quote. Amazon Athena charges per TB of data scanned by SQL queries (prices vary by region; use the AWS pricing page). Amazon EMR adds an EMR charge on top of the EC2 or other compute it runs on. IBM watsonx.data is priced by IBM. We have not quoted figures that the providers' pages load dynamically or by region.

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

Where each one leads

Trino strengths

Frequent releases and a broad documented connector list, including Iceberg, Delta Lake, Hudi and many databases
Fault-tolerant execution with task-level retries for long-running batch queries
Commercial and managed options from Starburst (Galaxy and Enterprise) and availability on Amazon EMR
Led by the original creators of Presto, under the non-profit Trino Software Foundation

Presto strengths

Governed by the Presto Foundation at the Linux Foundation, with corporate members
Presto C++ native workers on Velox, with no JVM on worker nodes
Offered as Java and C++ engines in IBM watsonx.data
Connectors for Iceberg, Delta Lake, Hudi and Hive, plus options such as Arrow Flight and Lance

Limitations

Trino limitations

Java only; there is no native C++ worker
Frequent releases mean regular upgrades and configuration changes to track
A query engine only: storage, catalogs and table maintenance are separate systems to run
Self-managed clusters need tuning for memory, concurrency and spilling

Presto limitations

Smaller commercial ecosystem; AWS recommends Trino over Presto on Amazon EMR
Presto C++ is in active development, with documented limits such as read-only Iceberg support
Shorter documented connector list than Trino for some systems, such as Snowflake and DuckDB
A query engine only, with the same operational overhead as Trino

When to choose each

Choose Trino if

You are starting a new federated query or lakehouse SQL deployment and want the more widely offered engine
You want a managed service or vendor support from Starburst
You need fault-tolerant execution for long-running batch queries
You need connectors that Trino documents and Presto does not, such as Snowflake or DuckDB

Choose Presto if

You want to use or contribute to Presto C++ and the Velox ecosystem
Your platform already standardises on Presto, for example IBM watsonx.data
Linux Foundation governance matters to your organisation
Your existing Presto deployment works and migration would add risk without a clear gain

When neither is right

You want a warehouse that stores and manages data for you rather than a query engine over other systems: see Snowflake vs BigQuery.
You need low-latency analytics served to applications on fresh event data: see ClickHouse vs Snowflake and ClickHouse vs BigQuery.
Your data fits on one machine: an embedded engine such as DuckDB is simpler; see DuckDB vs Snowflake.
You need a lakehouse platform with Spark, notebooks and machine learning: see ClickHouse vs Databricks and the best Databricks alternatives.

Final recommendation

Bottom line

Trino and Presto solve the same problem, SQL over data that lives in many places, and share an architecture and a history, but they are separate projects now. Trino is the default we would suggest for most new deployments because of its release pace, connector coverage, fault-tolerant execution and the managed and commercial options around it. Presto is the right choice when Presto C++ on Velox is part of your plan, when you run a platform built on it such as IBM watsonx.data, or when you already operate it successfully. Whichever you choose, test your own queries and connectors, because syntax and behaviour have diverged since 2020.

Frequently asked questions

Is Trino the same as Presto?

No longer. Trino began as PrestoSQL, a fork of Presto led by Presto's original creators, and was renamed Trino in December 2020. The two share an architecture and much of their SQL, but they are separate projects with different releases, features and governance.

Why was PrestoSQL renamed Trino?

According to the announcement on trino.io (27 December 2020), Facebook applied for a trademark on Presto and, after establishing the Presto Foundation at the Linux Foundation in September 2019, began enforcing it, so the PrestoSQL project changed its name to Trino.

Does Amazon Athena use Trino or Presto?

Athena engine version 3 is AWS's own engine. AWS says it integrates improvements from the open source Trino and Presto projects, and its documentation lists functionality incorporated from Trino and points to the Trino documentation for functions. Trino and Presto connectors are not supported in Athena; it uses Athena Federated Query instead.

What is Prestissimo?

Prestissimo is the development name for Presto C++, a C++ implementation of Presto workers built on the Velox library. It is a drop-in replacement for Java workers behind a Java coordinator. The Presto documentation describes it as in active development, with documented limits such as read-only support for Iceberg tables.

What is Starburst?

Starburst is a company founded by Trino's creators. It sells Starburst Galaxy, a managed service on AWS, GCP and Azure, and Starburst Enterprise for self-managed deployments, both built on Trino.

Sources

Checked October 2026.

How we research comparisons: our editorial method.