On July 9, 2026, OpenAI introduced "ChatGPT Work," a new agent inside ChatGPT built on Codex and GPT-5.6. It can take action across apps and files, stay on a single project for hours if needed, and turn a goal into finished work.
July 9, 2026 · OpenAI
ChatGPT Work: from chat replies to finished deliverables
A new agent built on Codex and GPT-5.6 that takes a goal, plans and executes on its own — staying on a single project for hours and returning documents, spreadsheets, presentations and reports.
1M+
token context for long-horizon reasoning
Hours
of autonomous, unattended work on one project
3
GPT-5.6 models: Sol · Terra · Luna
The shift in interaction model
One turn of chat vs. a full workflow, to scale
Previous ChatGPT
Single-turn reply
ChatGPT Work
Plan → execute → deliver
How it runs a goal
Give a goal
+ task document
→
Plan & act
across apps & files
→
Test until done
runs in background
→
Finished work
docs · sheets · decks
What users praise
Runs autonomously for hours, no micromanaging
Competitive analysis, financial schedules, decks
Code review, meeting coordination, spreadsheets
Where it still struggles
Usage and cost rise on large tasks
Misses linked clauses in complex 17-page contracts
Waiting / notification behavior needs refinement
Also launched the same day
Unified desktop app
Ties together Chat, Work and Codex
Scheduled Tasks
Recurring, trigger & change monitoring
Sites (beta)
Hosted sites generated as deliverables
Availability: paid ChatGPT plans first — Business, Enterprise & Edu — within existing Plus/Pro tiers.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…