Codex alpha adds Bedrock Ultrafast client support
Codex alpha adds client controls for Astra Ultrafast on supported Bedrock connections while preserving the default tier.
OpenAI published Codex 0.162.0 alpha 10 on October 3 at 09.55 UTC. The release is a developer preview, and its release page supplies only a version label.
The implementation change adds client support for selecting Astra Ultrafast through supported Amazon Bedrock connections. It also preserves service tiers defined in custom model catalogs.
For teams using Codex through AWS, the practical question is whether the client can pass the intended service tier to their endpoint. A setting in the terminal is one part of that path. Account access and the service receiving the request still matter.
A choice that leaves the default intact
The bundled catalog adds Ultrafast for Astra on Bedrock Mantle and the global and US Bedrock Runtime variants. It keeps the implicit default tier unless the user selects otherwise.
Custom catalogs retain their tier definitions and defaults. The client sends a selected tier only when that model advertises support, including for the Flex tier.
That boundary is useful when one application connects to several providers. A choice supported in one catalog should not silently become a request option everywhere else. Teams maintaining custom catalogs should check that their definitions accurately describe the endpoint they use.
What the source evidence establishes
Added tests cover catalog discovery, custom definitions, request handling and terminal selection. They check toggling Ultrafast and switching between Astra and a model without that tier.
ByteForward inspected the implementation and test code. We have not run this alpha binary or verified live AWS account access. The evidence establishes a client change, with no independent confirmation of service availability, pricing or performance.
Our earlier Codex alpha coverage examined Daybreak controls and custom provider settings for search and context. This update addresses the Bedrock service tier path.
Keep the trial separate from important work
For a controlled evaluation, record the exact build, provider, model and chosen tier. Confirm the outgoing request and returned service information against the provider documentation before drawing conclusions. Check that returning to the default setting works as intended.
This remains prerelease software. Test it away from work that would be costly to interrupt, and keep a known working setup available. The useful outcome is a verified configuration for a particular account and workflow.
Featured image is an original AI generated conceptual illustration of client service tier selection.
