OpenAI released five Codex CLI versions between September 17 and September 29, 2026. 0.155.0 introduced experimental voice conversations and Touch ID verification for MCP requests; 0.156.0 made voice default and added a fullscreen UI; 0.157.0 added GPT-6 Sol and Luna with Amazon Bedrock support; 0.159.0 added opt-in instant interrupt for steering mid-response. The cadence shows Codex maturing into a full-featured terminal agent with voice, multi-agent, and sandboxing infrastructure.
0.159.0 (2026-09-29) adds opt-in instant interrupt and Windows console-window fixes
0.156.0 promotes voice to default-on and adds fullscreen TUI and /usage analytics, building on 0.155.0's experimental voice rollout.
OpenAI released GPT-6 Sol and GPT-6 Luna on September 22, 2026, bringing GPT-6 Astra's advances in coding, factuality, computer use, and agentic work into faster, cheaper models. Sol costs $2/M input and $10/M output, Luna $0.10/M and $0.50/M — about 50% below GPT-5.6 promotional pricing. The launch came the same day as Anthropic's Claude Opus 5.5 and immediately reshaped the frontier pricing landscape.
OpenAI released GPT-6 Sol and GPT-6 Luna on September 22, 2026, bringing GPT-6 Astra's advances in coding, factuality, computer use, and agentic work into faster, cheaper models. Sol costs $2/M input and $10/M output, Luna $0.10/M and $0.50/M — about 50% below GPT-5.6 promotional pricing. The launch came the same day as Anthropic's Claude Opus 5.5 and immediately reshaped the frontier pricing landscape.
OpenAI released Codex CLI 0.154.0, adding GPT-6-Astra to the model picker and Amazon Bedrock catalogs, experimental git worktree support, inline mid-task questions, and a shared background Codex server for Windows sessions. A follow-up alpha, 0.155.0-alpha.1, shipped the next day.
OpenAI released Codex CLI 0.154.0, adding GPT-6-Astra to the model picker and Amazon Bedrock catalogs, experimental git worktree support, inline mid-task questions, and a shared background Codex server for Windows sessions. A follow-up alpha, 0.155.0-alpha.1, shipped the next day.
OpenAI released codex 0.161.0 on GitHub, promoting GPT-6.1 Sol to default in bundled and Amazon Bedrock catalogs. The release adds Bedrock multi-agent V2 and Ultra reasoning support, MCP server login from the terminal, voice device selection, and an opt-in Daybreak feature gating Cyber routing. Numerous bug fixes address permissions, Windows sessions, thread resume, and retry behavior.
OpenAI released codex 0.161.0 on GitHub, promoting GPT-6.1 Sol to default in bundled and Amazon Bedrock catalogs. The release adds Bedrock multi-agent V2 and Ultra reasoning support, MCP server login from the terminal, voice device selection, and an opt-in Daybreak feature gating Cyber routing. Numerous bug fixes address permissions, Windows sessions, thread resume, and retry behavior.
OpenAI released GPT-6.1 Sol at DevDay 2026, positioning it as a major upgrade over GPT-6 Sol that approaches GPT-6 Astra on coding, computer use, and professional work at one-fifth of Astra's standard API token prices. The model is available to Plus, Pro, Business, Enterprise, and Edu users in ChatGPT Work and Codex, and is generally available on Amazon Bedrock. OpenAI also launched a $500/month ChatGPT Pro 500 plan, and Simon Willison live-blogged the keynote.
OpenAI released GPT-6.1 Sol at DevDay 2026, positioning it as a major upgrade over GPT-6 Sol that approaches GPT-6 Astra on coding, computer use, and professional work at one-fifth of Astra's standard API token prices. The model is available to Plus, Pro, Business, Enterprise, and Edu users in ChatGPT Work and Codex, and is generally available on Amazon Bedrock. OpenAI also launched a $500/month ChatGPT Pro 500 plan, and Simon Willison live-blogged the keynote.
Amazon Bedrock now serves OpenAI's GPT-6 family (Sol, Luna, and Astra) through global cross-Region inference profiles, callable from Asia Pacific Regions in Singapore, Jakarta, Malaysia, Thailand, and Taipei. The AWS post walks through invocation via the Converse API, OpenAI-compatible APIs, and Codex CLI, plus IAM/SCP configuration, quota accounting, and monitoring. It matters to practitioners because single-Region inference can throttle when local capacity is exceeded, and enterprises commonly trip over SCP Region-allowlist denials on first calls.
Amazon Bedrock now serves OpenAI's GPT-6 family (Sol, Luna, and Astra) through global cross-Region inference profiles, callable from Asia Pacific Regions in Singapore, Jakarta, Malaysia, Thailand, and Taipei. The AWS post walks through invocation via the Converse API, OpenAI-compatible APIs, and Codex CLI, plus IAM/SCP configuration, quota accounting, and monitoring. It matters to practitioners because single-Region inference can throttle when local capacity is exceeded, and enterprises commonly trip over SCP Region-allowlist denials on first calls.