Issue #19: OpenAI just launched three new AI models. Here's which one your business should actually use.

It was delayed by a government review. It launched Thursday anyway. And it comes in three flavours with very different price tags.

On July 9, 2026, OpenAI publicly launched GPT-5.6 — its most capable AI model family yet. The rollout had been sitting in government review for two weeks, held back by the Trump administration after concerns were raised about its cybersecurity capabilities. The Department of Commerce cleared it for broad release after additional testing, and OpenAI moved fast.

For small business owners, most of the coverage has been technical and benchmark-heavy. This issue cuts through that. Here's what actually changed, and what it means for how you use ChatGPT day to day.

GPT-5.6 isn't a single model. It's a family of three, each built for a different job at a different price point. Understanding the difference saves you money and tells you what to actually use.

GPT-5.6 Luna — the everyday workhorse The fastest, cheapest model in the family. OpenAI says it outperforms Claude Opus 4.8 on coding benchmarks at roughly one-quarter the estimated cost. For small business owners using ChatGPT for emails, content drafts, summaries, and standard day-to-day tasks — Luna handles all of this comfortably. If you're on a Plus plan, this is what you'll mostly be running.

GPT-5.6 Terra — the balanced choice Terra performs at roughly the same level as GPT-5.5 — the previous best model — but at half the cost. For tasks that need a bit more reasoning — longer documents, complex analysis, multi-step planning — Terra is the right call without burning through tokens on Sol-level pricing. Free and Go users get Terra in ChatGPT's new Work mode.

GPT-5.6 Sol — the flagship OpenAI's most powerful model yet. Sol is described as their "strongest cybersecurity model ever" (which was precisely what got it held up in government review) and is built for complex reasoning, long-horizon agentic work, and high-stakes tasks. It scored 52.7% on the Agents' Last Exam benchmark — well above GPT-5.5's 46.9%. For most small business owners, Sol is overkill for daily use. But for deep research tasks, complex proposals, and long documents that need sharp reasoning, it's worth the extra tokens.

ChatGPT Work — the big new product Alongside the models, OpenAI launched ChatGPT Work — a new mode inside ChatGPT designed for complex, multi-step business tasks. In the launch demo, OpenAI's finance team showed it pulling data from Slack, running a variance analysis, updating an Excel model, generating a PowerPoint, and publishing a shareable dashboard — all from a single conversation. Free and Go users get Terra in Work mode. Paid users get Sol.

The verdict: For most small business owners, nothing dramatic needs to change. Luna handles everyday tasks at lower cost, and you'll get access to it automatically through your existing ChatGPT plan. The meaningful upgrade is ChatGPT Work for anyone who does complex, multi-step tasks regularly — the kind that previously required jumping between tools.

A small financial planning consultancy had been using ChatGPT Plus for client report generation — pulling data, writing analysis, formatting summaries. Their workflow involved at least four tools: ChatGPT for writing, Excel for numbers, Google Docs for formatting, and email for delivery.

The day GPT-5.6 dropped, they tested ChatGPT Work with a real client dataset. They uploaded a spreadsheet, described the report they needed, and asked it to pull the key metrics, write the analysis, and structure it as a shareable document. One conversation. One output. What previously took 90 minutes took 20.

They haven't abandoned their existing workflow yet — they're validating it over a few more reports first. But the direction is clear.

"Why did the government get involved in an AI model launch?"

GPT-5.6 Sol is, by OpenAI's own description, their strongest cybersecurity model yet — scoring dramatically higher than previous models at identifying software vulnerabilities and supporting offensive security tasks. The Trump administration signed an AI cybersecurity order in early June requiring companies to submit their most powerful models for government review 30 days before public release. OpenAI complied, shipped to a small group of approved partners on June 26, and received clearance for full public release in under two weeks after additional testing. This is worth understanding for a practical reason: the same export control logic that pulled Fable 5 offline last month was applied here, but resolved faster. Frontier AI model launches now involve government review as a real step in the process — which means more potential delays and more potential disruptions for businesses that build workflows around specific models.

For ChatGPT Plus subscribers: Sol is available through medium and higher effort settings, Terra for standard tasks, Luna for quick responses. You don't pay per token as a Plus subscriber — you pay the flat $20/month and the model selector determines which tier you access within your usage limits.

This week: open ChatGPT and find the new Work mode. Pick one multi-step task — something that currently takes you 30+ minutes and involves more than one tool. Run it through ChatGPT Work. You don't need to commit to changing anything. Just see how close it gets in one pass. That single test will tell you more about whether this matters for your specific workflow than any benchmark article will.

The GPT-5.6 launch happened the same week SpaceX's AI division launched Grok 4.5, and just days after Anthropic restored Fable 5 following its export control suspension. All three events happening within days of each other signals something meaningful: the frontier AI race has sped up to a point where major model launches, government interventions, and competitive responses are now happening on weekly timescales rather than quarterly ones. For small business owners, the practical implication isn't to chase every new model — it's to build workflows that can swap models without rebuilding from scratch.

Immediately switching everything to GPT-5.6 Sol just because it's the newest and most powerful option. Sol is built for complex, long-horizon tasks and frontier reasoning. It's overkill for drafting emails, generating social posts, or summarising documents — tasks that Luna and Terra handle perfectly well at significantly lower cost and usage impact. New model launches create an instinct to upgrade everything. Resist it. Identify one or two tasks where frontier reasoning would genuinely change the output quality. Use Sol there. Use Luna everywhere else.

Until next issue, Kris

The Layman's AI — The only AI updates your business actually needs.