Codex CLI: GPT-6 Sol and Luna Now Available via Amazon Bedrock
CodexOpenAI extended GPT-6 Sol and GPT-6 Luna to Amazon Bedrock, letting Codex CLI 0.157.0 users run both new model tiers through existing AWS accounts and billing instead of a separate OpenAI API key. The update, shipped September 25, 2026, three days after the models' original launch, also adds migration prompts nudging Bedrock users off GPT-5.6 Sol and Luna. It marks the fourth consecutive GPT model wave OpenAI has brought to Bedrock, reinforcing AWS as a standard secondary distribution channel for Codex-capable models.
Key Takeaways
- GPT-6 Sol and Luna are now available on Amazon Bedrock, three days after their original OpenAI API and Codex launch on September 22.
- Codex CLI 0.157.0 adds migration prompts nudging Bedrock users off older GPT-5.6 Sol and Luna models toward the GPT-6 tier.
- AWS customers can now run Codex through existing IAM, billing, and account infrastructure instead of managing a separate OpenAI API key.
- On Bedrock, GPT-6 Sol targets complex, recurring software development tasks while GPT-6 Luna handles high-volume, low-latency work like summarization and classification.
- Both models support context windows up to 1 million tokens on Bedrock, matching their direct-API specifications.
- This marks the fourth consecutive GPT model wave OpenAI has extended to Bedrock, following GPT-5.5, GPT-5.6 Sol/Terra/Luna, and the original April 2026 Bedrock integration.
Sources & Mentions
4 external resources covering this update
Bring more intelligence to everyday work with GPT-6 Sol and GPT-6 Luna on Amazon Bedrock
AWS Blog
OpenAI GPT-6 Sol and GPT-6 Luna are now generally available on Amazon Bedrock
AWS What's New
I tried out GPT-6 Sol/Luna, which became available on Amazon Bedrock
DevelopersIO
AWS AI: Spend intelligence where it changes the outcome
X
GPT-6 Sol and Luna Arrive on Amazon Bedrock
On September 25, 2026, Codex CLI 0.157.0 added GPT-6 Sol and GPT-6 Luna to its Amazon Bedrock model catalog, giving developers who run Codex through AWS infrastructure direct access to OpenAI's two newest mid- and low-cost model tiers. The update follows the models' original launch inside Codex, ChatGPT Work, and the OpenAI API on September 22, and extends the same pattern OpenAI has followed with every GPT-5.x and GPT-6 model wave: a direct-API rollout first, followed within days by general availability on Bedrock.
What Changed in Codex CLI
Codex CLI 0.157.0's release notes describe the change simply: "Added GPT-6 Sol and Luna, including Amazon Bedrock support and migration prompts for older models." In practice, this means AWS customers who configure Codex to use Bedrock as a model provider can now select gpt-6-sol and gpt-6-luna from the Bedrock console or invoke them programmatically through Bedrock's APIs, without needing a direct OpenAI API key. Codex also now surfaces migration prompts nudging users off older Bedrock-hosted models, such as GPT-5.6 Sol and Luna, toward the new GPT-6 tier.
Why Bedrock Access Matters
For organizations that standardize their cloud spending, IAM permissions, and billing through AWS, Bedrock access removes a significant adoption barrier. Rather than managing separate OpenAI billing and credentials, teams can run Codex against GPT-6 Sol or Luna using existing AWS accounts, budgets, and access controls. Amazon Web Services confirmed the general availability of both models on Bedrock in its own "What's New" and machine learning blog posts, noting that GPT-6 Sol handles "recurring complex tasks and software development" with roughly half the factual-error rate of GPT-5.6 Sol, while GPT-6 Luna targets high-volume, low-latency workloads like summarization and classification. Both models support context windows up to 1 million tokens on Bedrock.
Part of a Broader Pattern
This is now the fourth time OpenAI has extended a Codex-relevant model family to Bedrock, following GPT-5.6 Sol/Terra/Luna in July, GPT-5.5/Codex in June, and the original Bedrock integration in April. The consistency of this cadence suggests Bedrock has become a standard secondary distribution channel for every new Codex-capable model tier OpenAI ships, rather than a one-off partnership.