Codex CLI now defaults to GPT-6 Astra and adds it on Bedrock
Codex's bundled default is now Astra unless you set otherwise, it's selectable via your own AWS Bedrock routes, and Astra Fast is 2x speed — not 1.5x.
Copy markdown
Astra is Codex's new default
Codex 0.153.4 makes GPT-6 Astra the bundled model whenever you haven't set one explicitly. If you run Codex on autopilot you're now on Astra's $10/$50-per-Mtok rates by default — pin a cheaper model in your config if that's not the bill you want.
Route it through your own Bedrock
Version 0.153.3 added GPT-6 Astra to the Amazon Bedrock model picker for the Mantle and Runtime global/US routes. AWS teams can now point Codex at Astra through Bedrock, keeping requests inside their own cloud instead of calling OpenAI directly.
Astra Fast is 2x speed, not 1.5x
A text fix in 0.153.2 admits the Fast tier actually runs at “2x speed, increased usage” — the picker had been underselling it as 1.5x. Nothing about how requests run changed, so lean on Fast for tight iterative loops.
Elsewhere: GLM-5.3 is half-price on DigitalOcean
Vercel's changelog flags a 50% discount on GLM-5.3 via DigitalOcean running through Sept 7. If the open-weights “hyperscaler tax” scared you off self-hosting, this is a cheap managed window to benchmark it before the promo closes.