Neon

A September 4, 2026 post on the official Neon blog titled "Comparing token economics across 42 AI models" describes an experiment in which the Neon team ran a synthetic 100-support-ticket workload through models listed in the Neon AI Gateway and compared total cost, output usability, and time to completion, publishing The blog highlights specific results for developer decision-making: GPT-5 Nano was reported as the cheapest model for the task, cheaper than open-weight alternatives, with the lowest cost per usable response and completion in roughly three minutes; Llama 3.1 8B Instruct was the fastest at 1 minute 18 seconds but only p

Neon's official blog post by Bryan Clark (VP Product), originally published July 15, 2026, announced that Object Storage, Functions, and AI Gateway were entering beta and available to all Neon users. It describes object storage that branches with the database, long-running compute living next to the database, and an AI The post contextualizes the move by recalling Neon's 2022 bet to separate Postgres compute from storage and rebuild storage to make database environments lightweight and branchable — an architecture that turned out to fit agent workloads needing serverless economics and disposability. It positions the beta as the first

Changes.Watch, a third-party release-notes aggregator indexing Neon's official changelog, lists several recent Neon updates. The Sep 17, 2026 entry summarizes the Neon backend becoming generally available, offering a full serverless Postgres backend with branching support. The Sep 10, 2026 entry covers "Claimable Neon, An Aug 27, 2026 entry highlights three new Neon CLI commands to install Neon into coding agents such as Cursor, alongside new Neon backend guides and MCP stalled-query checks. Aggregate velocity figures on the product page record 1 update in the last 7 days, 4 in the last 30 days (3 features and 1 fix), 12 in the last

According to Neon's official Changelog (September 18, 2026 entry), the Neon backend is now generally available, expanding the platform from serverless Postgres into a full branchable backend. Every branch is a full copy, so a pull request, preview, or agent session gets the entire backend — schema, storage, auth, and f The same Neon Changelog entry walks developers through getting started with a coding agent via `neon skills`, the `neon deploy` workflow from a `neon.ts` file, and scaffolding options like `neon bootstrap`, linking to a full-stack quickstart and the GA launch blog post. It positions the platform as a coherent agent-dep