Automotive Grade Linux runs three concurrent release series on three Yocto baselines

One distribution, three baselines

Automotive Grade Linux (AGL) is not a single distribution on a single base. According to a post on dev.to by Raghu Bharadwaj published on 27 August 2026, the project's release-notes wiki — which records its own last edit as 24 August 2026 — listed three series as of 26 August:

• Vibrant Vimba, shown under the wiki's "Latest Stable Release" heading, built on Yocto Project Wrynose 6.0.x and covered by upstream support until April 2030. Its documented builds so far are Milestone 1 (v21.91.0) and Milestone 2 (v21.92.0).
• Ultimate Unagi, the 21.0.x series, built on Yocto Scarthgap 5.0.x, with upstream support running to April 2028. Its feature list includes the Scarthgap move, Qt 6.8, an upgrade to VSS 6.0, meta-flutter 3.38.3, the addition of meta-ros, KVM demo updates and Sparrow Hawk board support; point release v21.0.3 tracks Scarthgap 5.0.19 and adds SiFive HiFive Premier-P550 support.
• Quirky Quillback, the old-stable 17.x series, built on Yocto Kirkstone 4.0.x (4.0.35), which the Yocto Project now lists as end of life.

The post's wider point is that AGL's announcements, its product page and its release-notes wiki do not line up in August 2026: an engineer who starts from a press release can arrive at a different baseline than one who reads the release-notes table, and those baselines have upstream support windows that end two years apart.

Branch, manifest, baseline

The fetch commands reveal how differently the three series are pinned. Unagi and Quillback each have a cut release branch — repo init -b unagi or -b quillback — plus a pinned manifest such as unagi_21.0.3.xml. Vimba has no branch of its own: the wiki's command fetches master with a milestone manifest (vimba_21.92.0.xml). In other words, the series labelled the latest stable release rides on the line of active development, while Unagi offers both a manifest and a branch that is not moving underneath you.

The post also separates three ideas engineers often conflate: a branch is a moving line of development; a manifest pins a set of repository revisions and is what makes a build reproducible; a production baseline adds validation evidence and a maintenance plan naming who fixes what, and for how long.

Upstream windows, and their limits

Per the Yocto Project releases table, as cited in the post: Scarthgap 5.0 shipped in April 2024 with long-term support to April 2028; Wrynose 6.0 shipped in April 2026 with support to April 2030; Kirkstone 4.0 shipped in May 2022 and is now end of life. The contrast with non-LTS Yocto releases is sharp — Walnascar 5.2 and Whinlatter 5.3 both went end of life within roughly a year, and Blacksail 6.1, due October 2026, is scheduled for only six months of support.

The post qualifies these dates carefully: they are upstream Yocto horizons and nothing more. They say nothing about how long AGL itself will maintain a given series, how long a silicon vendor will support its BSP layers, or what a vehicle programme will require — three separate commitments owned by three separate parties, each of which has to be confirmed directly.

What moved between May and August

On 14 May 2026, at the AGL All Member Meeting in Tokyo, the Linux Foundation announced initial availability of the AGL SoDeV reference platform, delivered through the Ultimate Unagi Unified Code Base release. SoDeV combines the Unified Code Base with Linux containers, VirtIO, the Xen hypervisor and the Zephyr RTOS, targeting Renesas Sparrow Hawk boards, virtual machines and cloud environments. The announcement named Scarthgap 5.0.16 as the baseline, said Unagi would be supported for two years, and committed to updates roughly three weeks after each upstream Scarthgap release.

By the 24 August wiki edit, two things had shifted. Unagi had reached v21.0.3 on Scarthgap 5.0.19 — three point releases past the version named in May, visible in the artefacts rather than merely asserted — and Vibrant Vimba had taken over the "Latest Stable Release" heading on the newer Wrynose base.

Why it matters

Choosing a branch is a support-horizon decision, not a freshness decision. A platform entering production in 2028 and staying in service for a decade cannot rest on a base whose upstream fixes stop in April 2028, and even the April 2030 horizon only defers the security-backport burden by two years before that work lands on your own team — the same arithmetic that governs kernel selection. Quirky Quillback stays listed so that programmes validated years ago can still reproduce their builds; anyone fetching it today takes on the backporting themselves. And because upstream windows say nothing about AGL's own maintenance or vendor BSP commitments, every layer of support has to be verified with the party that actually provides it.