OpenAI Codex July 13: CLI 0.144.x Adds Usage-Limit Reset Credits and a writes Approval Mode, GPT-5.6 Sol/Terra/Luna Land, Codex Folds Into the ChatGPT Desktop App

OpenAI Releases GPT-5.6 Sol, Terra and Luna: Three-Tier Family Live in ChatGPT, Codex and the API, Programmatic Tool Calling Cuts Tokens 38-63.5%

• On July 9, OpenAI made the GPT-5.6 family generally available across ChatGPT, Codex and the API.
• Sol is the flagship for frontier reasoning and long-horizon agentic work ($5 input / $30 output per 1M tokens), Terra is the balanced everyday model ($2.50 / $15), and Luna is the fastest and cheapest ($1 / $6).
• New Programmatic Tool Calling lets GPT-5.6 write and run in-memory JavaScript to coordinate tools, cutting tokens 38-63.5% in OpenAI's tests.
• Sol leads the Artificial Analysis Coding Agent Index at 80 but trails Claude on SWE-Bench Pro by about 15 points.
• GPT-5.4 retires July 23.

Editor’s read: Pick your tier deliberately: Terra matches GPT-5.5 at roughly 2x lower cost, and Programmatic Tool Calling is a real token saver for tool-heavy agent runs. GPT-5.4 is gone July 23, so migrate pinned configs now.

Codex CLI 0.144.x: Usage-Limit Reset Credits, a New writes Approval Mode, Interactive MCP Auth - Plus 0.144.1/0.144.2 Install and Guardian Fixes

• On July 9, Codex CLI 0.144.0 shipped usage-limit reset credits (with type and expiration shown, and a choice of which to redeem), a new writes app-approval mode that allows declared read-only actions while prompting for writes, interactive MCP authentication, and an Ultra high-concurrency warning.
• 0.144.1 (July 9) fixed standalone installs failing on compact GitHub release metadata and exposed the macOS code-mode host.
• 0.144.2 (July 13) restored the previous Guardian auto-review policy and tool behavior after rolling back a prompting regression.

Editor’s read: The writes approval mode is the practical win here: read-only actions run freely while writes still prompt. And if you saw odd auto-review behavior mid-week, 0.144.2 rolls it back - update to it.

OpenAI Codex CLI v0.143.0: Remote Plugins On by Default, macOS/Windows System Proxy Support, Amazon Bedrock GPT-5.6 Models

• On July 8, Codex CLI v0.143.0 turned remote plugins on by default with npm marketplace sources and visible remote/local versions, routed authentication and Responses API traffic through macOS and Windows system proxies (including PAC and WPAD), added Amazon Bedrock GPT-5.6 Sol/Terra/Luna models with max reasoning effort, and made MCP tools use tool search by default.

OpenAI Folds Codex Into the Unified ChatGPT Desktop App (macOS, Windows): Sidebar GitHub PR Review, Multi-Repo Projects, ChatGPT Work Targets Claude Cowork

• On July 9, OpenAI merged the standalone Codex app into the new unified ChatGPT desktop app for macOS and Windows.
• Codex stays a dedicated coding experience next to Chat and Work: you can now review GitHub pull requests in the sidebar alongside the diff and work across multiple repositories in a single project.
• The companion ChatGPT Work agent - powered by GPT-5.6 and pulling from Slack, Teams, Google Drive and CRMs - competes directly with Anthropic's Claude Cowork.
• Codex now has more than 5 million weekly active users, up 6x since the February desktop launch.

Editor’s read: The sidebar PR review inside the desktop app is the feature most Codex users will feel day to day - it collapses the review loop into one window.

Apple Sues OpenAI for Trade Secret Theft: Complaint Names Codex Maker's Hardware Chief Tang Tan, Filed in Northern California July 10

• On July 10, Apple sued OpenAI in the Northern District of California alleging trade secret theft and breach of contract, naming OpenAI Chief Hardware Officer and 24-year Apple veteran Tang Tan as directing the misconduct, and alleging a former Apple engineer downloaded confidential documents before joining OpenAI.
• OpenAI said it has no interest in other companies' trade secrets.
• Relations chilled after OpenAI bought Jony Ive's IO Products for $6.4 billion and moved into consumer hardware.

Best Model for Coding After a Chaotic Week: GPT-5.6 vs Grok 4.5 vs Claude Fable 5 on Price and Benchmarks

• On July 10, Yellow.com compared the week's releases: GPT-5.6 Sol lists at $5/$30 per million tokens, Grok 4.5 at $2/$6, and Claude Fable 5 at $10/$50.
• Fable 5 leads coding at 80.4% on SWE-Bench Pro versus Grok 4.5's 64.7% and GPT-5.5's 58.6%, while independent analysis shows Grok 4.5 completing comparable coding tasks for $2.49 versus $11.80 for Fable 5.