Research · July 13, 2026 · 10 min read
ChatGPT Work + GPT-5.6: What OpenAI Just Announced for Small Businesses
On July 9, 2026, OpenAI shipped two things at once that reshape what “ChatGPT for business” means: GPT-5.6 in three performance-cost tiers (Sol, Terra, Luna) and a new agent named ChatGPT Work that blends OpenAI's Codex coding tool with ChatGPT to take action across your apps and files.
For SMB owners, this is the biggest OpenAI announcement since the May 5 GPT-5.5 default rollout. It also completes a three-way race with Microsoft 365 with Copilot (launched July 1) and Claude for Small Business (launched May 13) — three frontier labs, three SMB-direct products, each betting on a different form factor. This post covers what's actually new (verified where possible, honest about sourcing where not), how it compares, and what to do this week.
What ChatGPT Work actually does
The best short description from the coverage: “an agent built for everyday work that blends Codex + ChatGPT to handle longer tasks, take action across your apps and files, stay with a project for hours if needed, and turn a goal into finished work.”
In practice, that means: a user asks ChatGPT Work to do a multi-step task — build a spreadsheet, draft a presentation, write a simple web app, or produce a document by pulling from multiple sources — and ChatGPT Work delivers the finished output, not just an answer.
Reported deliverables it can produce today:
- Documents
- Spreadsheets
- Presentations
- Simple web apps
The unified plugins directory
ChatGPT Work introduces a unified plugins directory. Named integration partners at launch, based on public reporting:
Slack
Gmail
Google Drive
Calendars
CRMs
The plugins directory is what makes the “take action across your apps and files” claim concrete. When you ask ChatGPT Work to “pull last quarter's expense receipts from Gmail, categorize them, and produce a summary spreadsheet,” the plugins directory is the connection layer.
GPT-5.6 in three tiers — Sol, Terra, Luna
OpenAI also shipped GPT-5.6 in three named tiers. The number (5.6) identifies the model generation; the name (Sol / Terra / Luna) identifies a durable capability tier that can advance on its own cadence — meaning future updates can lift each tier independently.
Sol
Top tier — most capableAPI price
$5 / $30 per 1M tokens (input / output)
Use for
Complex reasoning, agentic tasks, long-context work. ChatGPT Work uses Sol for its highest-complexity operations.
Terra
Mid tier — balancedAPI price
$2.50 / $15 per 1M tokens (input / output)
Use for
Everyday business work, drafting, analysis. The default choice for most SMB tasks that need more than a fast response but don't require Sol's full power.
Luna
Fast tier — cheapestAPI price
$1 / $6 per 1M tokens (input / output)
Use for
High-volume, quick-response work. Best for classification, short summaries, simple Q&A at scale. Compare with prior Flash-tier pricing from other providers.
Note: OpenAI's own release-notes page returned 403 at fetch time. Prices above are consistent across multiple secondary outlets covering the July 9 launch; verify against your OpenAI billing dashboard before major architectural decisions.
Who gets ChatGPT Work first
Reported rollout order for ChatGPT Work on web and mobile (Free and Go tiers do not get it):
- First wave (July 9): ChatGPT Pro, Pro Lite, Enterprise, and Edu
- Second wave (coming days): ChatGPT Plus and Business
- Not included: Free and Go tiers (at launch)
For SMBs on ChatGPT Business, this means the agent should start appearing in your dashboard this week if it hasn't already. If you don't see it yet, the reported explanation is a staged rollout: it depends on your plan, region, and timing.
How ChatGPT Work compares to the alternatives
Three products, all shipped within about two months:
ChatGPT Work + GPT-5.6
Launched
July 9, 2026
Strength
Best model access — Sol is at the frontier. Blends Codex + ChatGPT so it can build simple web apps, not just documents. Unified plugins directory. Doesn't require you to change productivity stack.
Limit
ChatGPT-native experience, not embedded in existing tools like Word/Sheets. Requires ChatGPT Business/Plus/Pro; Free tier excluded. Best-in-tier price ($25/user for ChatGPT Business) but token usage on top for API-heavy workloads.
Microsoft 365 with Copilot
Launched
July 1, 2026
Strength
Deepest integration with Word, Excel, PowerPoint, Outlook. Multi-model (OpenAI + Anthropic). Built-in DLP. 1,000+ connectors. Best if your team is already on M365.
Limit
Only inside Microsoft 365 apps. Requires the productivity stack. Pricing not disclosed at launch — check with your Microsoft Partner.
Claude for Small Business
Launched
May 13, 2026
Strength
Deep back-office operational connectors (QuickBooks, PayPal, HubSpot). 15 pre-built agentic workflows for finance, HR, ops, marketing. Anthropic-only, so single-model consistency.
Limit
No e-commerce platform connectors at launch. Not embedded in Word/Sheets. Anthropic-only means no multi-model fallback.
The decision framework — updated for July
If you already have ChatGPT Business, wait for ChatGPT Work to appear this week
You've already paid for the tier that gets it in the second-wave rollout. Nothing to buy. Try it on one real deliverable — say, next week's team update slide deck built from Slack messages and Google Drive files. Measure the time saved.
If you're on Microsoft 365, stay on M365 + Copilot
The July 1 M365 Business Copilot launch is still the right pick if your team is deep in Word/Excel/Outlook. ChatGPT Work's advantage is the agent + Codex hybrid; M365 Copilot's advantage is the productivity-app depth. Different needs.
If your bottleneck is finance / contracts / back-office, Claude for Small Business is still the specialist pick
Neither ChatGPT Work nor M365 Copilot has QuickBooks + PayPal + HubSpot connectors and 15 pre-built agentic workflows for those specific use cases. Claude for Small Business is unmatched in that lane.
Ignore the model-tier questions until the workflow question is answered
Sol vs Terra vs Luna matters for API-heavy applications. For most SMB use cases through the ChatGPT UI, the default tier is fine — ChatGPT Work will route to the appropriate model. Don't overthink model selection when workflow fit is the bigger unknown.
The bigger picture
Between May 13 (Claude for Small Business), May 19 (Google Gemini for Workspace + Spark), July 1 (M365 Business with Copilot), and July 9 (ChatGPT Work), the frontier AI labs have collectively launched four SMB-direct AI products in under two months. The competitive intensity is real. For SMB owners, that means pricing and features will keep improving — but it also means the “which AI is right for us” question is now permanent, not a one-time decision.
The stable answer: pilot the one that fits your current productivity stack, measure ROI on one real workflow, and re-evaluate every 6 months. The tools are moving faster than any single-vendor lock-in strategy can keep up with.
The AI-inside-your-business decision is separate from the AI-finds-your-business decision
ChatGPT Work vs M365 Copilot vs Claude is a productivity question. Whether customers can find your business across ChatGPT, Claude, Perplexity, Google AI Mode, and Siri is a visibility question. Our free scan handles the second. Under 10 seconds, no signup.
Run my free scan →Sources: OpenAI — Introducing ChatGPT Work (page returned 403 at fetch time; content confirmed via multiple secondary outlets), Axios — OpenAI releases GPT-5.6 and ChatGPT Work, TestingCatalog — OpenAI launches GPT-5.6 Sol, Terra, Luna, AI Weekly — OpenAI ships GPT-5.6 and ChatGPT Work. Where facts differ across sources, we've deferred to the majority-consensus version and flagged this uncertainty explicitly.
Written by the team at Kesem Marketing, a digital agency helping small businesses get found in the AI-first era.