XNews

OpenAI ChatGPT Work and GPT-5.6 Product Suite Launch

OpenAI simultaneously released GPT-5.6, merged ChatGPT and Codex into a single desktop application, and introduced ChatGPT Work, an agentic capability that executes multi-application tasks autonomously. Product architecture: • Three modes in one app: Chat (conversational AI), Work (autonomous agent connecting business tools), and Codex (coding agent). Former Codex app merged into ChatGPT desktop; optional Codex icon and default view. • GPT-5.6 model family: Sol (flagship, $5/$30 per million tokens), Terra (balanced, $2.50/$15), Luna (fastest/cheapest, $1/$6). Sol supports max and ultra reasoning levels; ultra splits tasks across sub-agents, restricted to Pro/Enterprise. • ChatGPT Work: uses plugins to connect Slack, Gmail, Drive, calendars, CRM, etc.; operates autonomously for hours, breaks down complex projects, requests human input only when needed. Supports scheduled tasks (one-time, recurring, event-triggered, monitoring). Runs on web, mobile, and desktop with local file access and a built-in browser. • Agentic usage is a separate quota pool shared by Work, Codex, ChatGPT for Excel, and Workspace Agents, distinct from Chat messages. All paid plans include Work; Free tier gets very limited access. • Pricing: Free (limited, US ads), Go ($8/mo), Plus ($20/mo, full features), Pro ($100/$200), Business ($20/user/mo), Enterprise. • GPT-5.4 retires July 23, 2026. Codex includes a Claude Code migration tool. • New Sites feature: publish any chat-generated visualization, dashboard, or tool as a shareable website. • Browser enhancements: session-aware tabs, file downloads, Chrome extension for direct action. OpenAI described the release as part of a super-app strategy, targeting enterprise productivity ahead of its IPO.

Key insights

The Unified App and ChatGPT Work Signal a Platform Play Against Office Suites

By combining a conversational assistant, an agent that operates across business apps, and a coding agent into one application with shared context, OpenAI positions itself as a direct competitor to Google Workspace and Microsoft 365. The ability to schedule recurring tasks and publish live Sites from conversations extends the offering from an occasional tool to a persistent productivity layer.

Agentic Usage Pooling Forces Users to Trade Off Between Coding and Business Automation

Work and Codex share a single 'agentic usage' pool, separate from Chat limits. This means heavy Work usage in one project can deplete quota for Codex coding sessions later. Users on lower plans will need to manage this trade-off carefully, potentially pushing heavy users toward Pro or Enterprise tiers.

The GPT-5.6 Tiered Model Strategy and Ultra Mode Extend Premium Differentiation

Offering three capability/cost levels and an ultra reasoning mode that invokes parallel sub-agents creates a clear upgrade path. This locks the most advanced agentic features behind Pro/Enterprise, incentivizing enterprise adoption while giving casual users a taste via Terra on lower plans. The retirement of GPT-5.4 accelerates migration.

Codex's Claude Code Migration Tool Targets Anthropic Developers Directly

The inclusion of a Claude Code import feature, combined with a unified desktop experience, is a tactical move to attract developers currently invested in Anthropic's ecosystem. Combined with the broader product suite, it aims to make the OpenAI platform the default for both technical and non-technical agentic tasks.

Source posts