# OpenAI Launches GPT-5.6 and ChatGPT Work: The Biggest ChatGPT Overhaul Yet By The Current Tribune · Technology · Published Sat, 11 Jul 2026 05:18:24 GMT · Updated Sat, 11 Jul 2026 11:18:24 GMT Source: The Current Tribune — https://currenttribune.com/article/openai-gpt-5-6-chatgpt-work OpenAI just dropped its most sweeping ChatGPT update to date — and it’s less a single product launch than a full reset of what ChatGPT is supposed to be. On July 9, 2026, the company rolled out the GPT-5.6 family of models, a brand-new agent called ChatGPT Work, a unified desktop app with coding tool Codex baked in, a public beta for building live websites, and a fresh naming scheme built around three model tiers named Sol, Terra, and Luna. If you’ve felt like OpenAI’s product lineup had become a confusing sprawl of chatbots, coding agents, and browsers, this release is the company’s attempt to fold it all into one place. Here’s everything OpenAI announced, what’s actually new, and who gets access first. ### Meet Sol, Terra, and Luna: A New Naming System The headline is GPT-5.6, but the more interesting change is how OpenAI is now labeling its models. Instead of piling on version numbers and confusing suffixes, the number now identifies the generation, while three names identify durable capability tiers that can each advance on their own schedule. - **Sol** is the flagship — OpenAI’s most capable model yet across coding, knowledge work, cybersecurity, and science. - **Terra** is the balanced everyday workhorse, delivering performance comparable to the older GPT-5.5 at roughly half the cost. - **Luna** is the fastest and cheapest option, aimed at high-volume, cost-sensitive tasks. OpenAI’s pitch throughout the launch is “performance per dollar” — the idea that GPT-5.6 squeezes more useful output from every token. Sam Altman told CNBC that Sol is 54% more token-efficient on agentic coding tasks than its predecessor, a stat clearly aimed at enterprises now scrutinizing every dollar they spend on AI. #### Max and Ultra: Two New Reasoning Modes For heavier jobs, GPT-5.6 introduces two escalating effort settings. **Max** hands the model more time to reason, run checks, and revise its approach. **Ultra** goes further, coordinating four agents in parallel by default to attack complex problems faster — with 16-agent configurations shown in some benchmarks. In ChatGPT Work, Ultra is reserved for Pro and Enterprise users; in Codex, it’s available on Plus plans and above. ### GPT-5.6 Pricing and Benchmarks On paper, the numbers are strong. OpenAI says Sol sets a new state-of-the-art on the Artificial Analysis Coding Agent Index at 80 and hits 53.6 on Agents’ Last Exam, a test of long-running professional workflows. It also claims a significant jump in cybersecurity, scoring 73.5% on ExploitBench compared to GPT-5.5’s 47.9% under a comparable token budget. Here’s how the three tiers compare on price, billed per one million tokens: Model Input (per 1M) Output (per 1M) Best for GPT-5.6 Sol $5.00 $30.00 Complex, high-stakes work GPT-5.6 Terra $2.50 $15.00 Balanced everyday tasks GPT-5.6 Luna $1.00 $6.00 Fast, budget-friendly jobs Developers get two more toys through the API: **Programmatic Tool Calling**, which lets the model write and run small programs to coordinate tools and filter data without bouncing every response back through the model, and a **Multi-agent beta** that runs several subagents at once and synthesizes their work into a single answer. ### ChatGPT Work: OpenAI’s New AI “Co-Worker” The star of the show is ChatGPT Work, a new agent that reframes ChatGPT from a chatbot into more of a digital colleague. Powered by GPT-5.6 and Codex, it’s designed for longer, multi-step assignments rather than one-off prompts — the kind of work that unfolds over hours. #### What ChatGPT Work Can Actually Do Point it at a goal, and ChatGPT Work gathers context from your connected apps and files, then produces finished artifacts: spreadsheets, slide decks, documents, and even web apps. It breaks big projects into smaller steps and keeps grinding independently in the background. OpenAI says nearly 100% of its internal teams now use the agent, with finance teams reportedly cutting their month-end close from days to hours. #### Plugins, Scheduled Tasks, and Sites Three supporting features make the agent genuinely useful: - **Plugins** connect ChatGPT to the tools where work already happens — Slack, Microsoft Teams, Google Drive, SharePoint, email, calendars, CRMs, and project trackers. Type “@” to pull context from a specific app, or let ChatGPT decide when to reference one. - **Scheduled Tasks** automate recurring work — reviewing weekly Slack updates, refreshing a meeting agenda, checking dashboards each morning, or updating a presentation when new feedback lands by email. - **Sites** (now in public beta) turns your information into interactive websites and web apps like dashboards, project trackers, and internal portals — and keeps them fresh as the underlying data changes. For enterprise buyers, OpenAI is leaning on governance: a Compliance API for visibility into Work conversations and actions, admin controls over connected tools and browser permissions, and an auto-review layer that vets sensitive actions before they run. ### One Desktop App to Rule Them All The other big structural change is on the desktop. OpenAI has merged its standalone Codex coding app into a single, unified ChatGPT desktop app for Mac and Windows, so you can switch between regular chat, ChatGPT Work, and Codex from one window. Existing Codex users keep their projects and settings after an update, and the old ChatGPT desktop app is being renamed to ChatGPT Classic. That app now packs a **built-in browser** for pulling in websites and online files, plus **Computer Use** — the ability to click, type, and move files across your local apps with your approval, either as a one-off or as part of a scheduled task. OpenAI also refreshed its Chrome extension so ChatGPT can work from Chrome’s sidebar. The casualty of all this consolidation is **Atlas**, OpenAI’s experimental agentic browser, which is scheduled to stop working on August 9, 2026. The company says the lessons from Atlas are moving into ChatGPT itself. If you use Atlas, back up your bookmarks, tabs, and history before then — that data won’t transfer automatically. ### GPT-Live: A More Human Voice Announced a day earlier, **GPT-Live** is a new generation of voice models replacing the old ChatGPT Voice experience. It can listen and speak at the same time, making interruptions and turn-taking feel more natural, and even throws in acknowledgment sounds like “mhmm” to show it’s paying attention. Paid users get GPT-Live-1, while free users get a mini version, rolling out across web, iOS, and Android in supported regions. ### The Government Delay and the Anthropic Rivalry There’s a political backstory here. OpenAI first previewed GPT-5.6 in late June but held the public release after the Trump administration — citing a June AI cybersecurity executive order — asked it to limit early access to a small group of government-approved partners. After additional testing by the Commerce Department’s Center for AI Standards and Innovation, OpenAI got the green light for a wide launch. Altman said the company made “many changes” after a collaborative back-and-forth, though it made clear it doesn’t want government preclearance to become the norm. The launch also lands squarely in OpenAI’s ongoing brawl with Anthropic for the “most powerful model” crown. Early reactions are mixed: some testers rave about GPT-5.6, while others say Anthropic’s Fable 5 still edges it out on many tasks. OpenAI’s counterargument is efficiency — matching or beating rivals while burning far fewer tokens and dollars. ### Who Gets What, and When Availability is staged. On desktop, Chat, Work, and Codex are live globally on Mac and Windows across all plans, including Free. ChatGPT Work on web and mobile rolled out first to Pro, Enterprise, and Edu users, with Plus and Business following within days. Free and Go users get Terra inside Work and Codex, while paid tiers can pick among Sol, Terra, and Luna. One housekeeping note: OpenAI confirmed that GPT-5.4 will retire on July 23, while GPT-5.5 sticks around. ### Final Verdict GPT-5.6 is a solid, efficiency-focused upgrade, but it’s the surrounding package that makes this release matter. By merging Codex into a single desktop app, launching a genuine work agent, and sunsetting Atlas, OpenAI is finally addressing the criticism that its lineup had become a tangle of overlapping products. ChatGPT Work is the clearest signal yet of where the company is headed — away from the chatbot and toward an autonomous agent that ships finished spreadsheets, decks, and dashboards while you’re away from your desk. The caveats are real. The most impressive productivity numbers come from OpenAI’s own internal teams and early partners, not from independent testing, and granting an agent computer access to your local files is a security decision worth taking seriously. Whether GPT-5.6 truly out-muscles Anthropic’s best is still up for debate. But if you use ChatGPT for actual work, this update turns it from a clever assistant into something that behaves much more like a co-worker. Treat it as an experiment first — approvals on, expectations measured — and it could genuinely change how your week runs.