Supabase 2026 changelog, condensed. What actually shipped, with links back to the source.

Why this page exists

The Supabase team ships fast. The changelog reflects that, which is a good problem to have and a bad reading experience. Most weeks contain a mix of dashboard polish, deeper platform changes, SDK bumps, and beta-to-GA promotions. For a working developer, only two of those four usually matter. The goal here is to flatten the noise and surface the items that actually affect production code: version bumps, breaking API changes, beta-to-GA graduations, pricing changes, and platform availability shifts.

Everything below is paraphrased from the public changelog. If I have an opinion on a release, it is marked as such. If you need to cite something for an internal doc or a customer email, link to the original changelog post, not to this page.

The shape of the 2026 changelog so far

Five themes have dominated 2026 releases through mid-May. Most individual posts on supabase.com/changelog fall under one of these.

January through March: platform foundations

The first quarter was heavy on infrastructure work that is invisible until you need it. Postgres 17 went from preview to default for new projects. pgvector picked up the HNSW improvements that landed upstream, which most teams will notice as a small drop in index build time and a modest accuracy bump at the default parameters. The Foreign Data Wrapper for Iceberg shipped, which is the meaningful one if you have analytics data sitting in object storage and want to query it from Postgres without copying.

The dashboard also picked up a redesigned table editor with a notably faster paginator on tables with millions of rows. Small change, real impact for anyone doing manual data audits in production.

April: Queues and Cron leave beta

Both moved to GA in the same window. Cron is a UI plus REST surface over pg_cron, which means every scheduled job is a row in cron.job and every run is a row in cron.job_run_details. That gives you SQL access to job history out of the box, which most managed cron offerings hide behind a separate API.

Queues is built on pgmq and exposes typed message payloads, visibility timeouts, and a poll-and-acknowledge flow. The performance ceiling is your Postgres write throughput, which is fine for most application workloads and not fine for ten-thousand jobs per minute. The honest reading: it removes the “do I need Redis just for this one job queue” question for the long tail of small apps.

May: Realtime Broadcast bypass and Edge Function regions

Realtime got two changes that compound. The bigger one is the WebSocket-only Broadcast path, which skips the round trip through Postgres for messages that you do not need to persist. Cursor movements, typing indicators, and live-presence dot positions all fit. The smaller one is a presence payload that is roughly half the size on the wire, which is meaningful for rooms with many subscribers.

Edge Functions picked up region pinning. Practical effect: a function that reads from a same-region Postgres instance is now consistently fast in tail latency, where in 2025 the global edge sometimes routed a request through a far region and added a hundred milliseconds. If you have a function that reads from the database on every call, pin it to the same region as your DB.

How to verify a release affects your project

A changelog post is a claim about the platform. Whether it has reached your specific project is a separate question. Two checks remove most of the ambiguity.

The first command returns the Postgres version the project is actually on, which is the most common “did this reach me” question. The second pulls the raw changelog feed so you can grep for the feature name and timestamp. If a feature is visible in the feed but not in your project, you are usually waiting on a rolling deploy or an opt-in flag.

The releases I would not skip reading

If you only have time for a handful of changelog posts, these are the ones that are most likely to change how you write code against Supabase in 2026.

• Postgres 17 default and the extension compatibility list. The dashboard upgrade page tells you which extensions need a rebuild. Read it before you upgrade a prod project.
• Queues GA. If you maintain a worker pool reading from Redis or SQS today, do the cost math on moving short, durable jobs to Queues.
• Realtime channel authorization via RLS. The migration is mechanical but it is not a no-op. Existing channel-level tokens still work; new code should use the policy-based path.
• Branching seed scripts. Check seeds into the repo, run them on branch create, and your preview environments stop being snowflakes.
• Edge Function region pinning. A two-line config change in many cases, and the single fastest way to drop tail latency on read-heavy functions.

Common questions