Google updates A2UI with swappable inference formats and Express mode
In short: Google has updated its A2UI generative UI protocol to include swappable inference formats, addressing latency and reliability issues found in the previous Direct JSON approach. A new format, A2UI Express, reduces token usage by up to 3x and latency by up to 3x by using a concise Python-like syntax.
This summary was generated automatically by AI from Google Developers's publication. It is our own text, not a copy of the original — facts, figures and quotes belong to the source, linked above and below.
What changed?
- 1Introduction of swappable inference formats: Express, Atom, Elemental, and custom.
- 2A2UI Express becomes the default, offering a Python-like syntax that improves inference speed and accuracy.
- 3New SDK support in Python (a2ui-agent-sdk 0.9.0), TypeScript, and Dart.
- 4Support for 'Controlled Generation' using decision models like Jev for even faster, layout-specific UI construction.
| Parameter | Before | Now |
|---|---|---|
| Inference format | Direct JSON | Swappable (Express, Atom, Elemental, or custom) |
| Token efficiency (Express vs Direct JSON) | Not specified | Up to 3x fewer tokens |
| Latency reduction (Express) | Not specified | Up to 3x lower |
Why it matters
For developers building agentic UIs, these changes allow for a significant reduction in latency and token costs. By decoupling the model output from the transport protocol via compilers, agents become more reliable and easier to integrate with specialized decision-making models.
What it means for AI agents and contact centers
If you are scaling complex agent-driven interfaces, switching to A2UI Express should be a priority for performance optimization. When testing, benchmark the latency gains against your existing Direct JSON implementations and evaluate if your specific screen layouts could benefit from the 'controlled generation' pattern with Jev.
🧪 Worth Testing
The 3x reduction in latency and token usage provides a clear advantage for production-grade agentic UIs.
Sources
- Google DevelopersOfficialPrimary sourceOriginal article →„Fast, reliable A2UI, with swappable inference formats“27 Oct 2026, 02:00
- Published by source
- 27 Oct 2026, 02:00
- Found by our system
- 27 Oct 2026, 17:22
- Summary generated
- 27 Oct 2026, 17:22
This article was written by AI from the original source. Facts, numbers and prices come from the source; missing values are marked “Not specified”. Legal notice, copyright and privacy