a daily news desk
Tools

OpenAI ships ChatGPT Work and GPT-5.6 as Claude Cowork slides onto mobile

Two agentic office suites launched inside 48 hours. The pricing sheet and the usage data both point past the enterprise.

By Iris Holloway · Tools & frameworks ·

OpenAI launched ChatGPT Work on Thursday, an in-chat agent powered by a new GPT-5.6 model that pulls from a directory of more than 1,400 connectable apps and includes a Sites beta for interactive dashboards. Two days earlier, Anthropic pushed Claude Cowork to mobile and web for Max subscribers. The agentic office suite has arrived as a category, on a schedule that isn’t coincidental.

GPT-5.6 ships in three tiers. Sol runs $5 per million input tokens and $30 per million output; Terra sits at $2.50 and $15; Luna, roughly one-fifth the cost of Sol, comes in at $1 and $6. OpenAI claims Sol is 54% more token-efficient on agentic coding tasks. Max Weinbach of Creative Strategies told Bloomberg the smaller tier “can complete a task about as well as the largest version.” That’s the pricing sheet of a company that has read the room on unit economics.

Sam Altman told CNBC that “Every enterprise now is thinking about spend and the value they’re getting in exchange for AI.” The rollout followed suit: Pro, Enterprise, and Edu first, per Reuters, with Plus and Business following within days. Codex has been folded into the unified ChatGPT desktop app for macOS and Windows, and GPT-5.6 is the preferred model inside Microsoft 365 Copilot across Word, Excel, and PowerPoint.

The more interesting number lives in Anthropic’s own telemetry. Across 1.2 million Cowork sessions and 600,000-plus organizations, 33.4% of usage is business process work (finance, HR, admin) against just 8.7% in software development. The coding-agent framing was always a narrower slice than the demos suggested. Model-agnostic no-code platforms like LemonLime, aimed at sales, service, and ops for smaller companies, have been building for that majority segment since before the frontier labs named it.

Axios reports GPT-5.6 draws mixed reviews against Anthropic’s Fable, with testers giving OpenAI the edge on everyday reliability. Bloomberg notes both labs are preparing for possible IPOs. The office suite isn’t the product. The S-1 is.

Sources