OpenAI launches dots, Sol matches Sonnet's price, Claude naps

On September 29 at DevDay, OpenAI introduced dots, always-on agents powered by GPT-6 Astra. Each dot gets its own cloud computer and browser, connects to more than 4,000 apps through plugins, and takes messages in ChatGPT, Slack and Teams, with texting coming soon. Dots are rolling out to Pro and Business Premium subscribers in eligible markets, and Enterprise workspaces can try a beta once an admin switches it on. Your first dot is included in the plan at no extra cost. You name it, so choose well: you will be typing that name into Slack a lot.

The marketing examples are OpenAI's own. One dot revises launch materials when a product's scope changes. Another takes a new interview transcript and comes back with clip ideas, show notes and social posts drafted for approval. OpenAI says it tested specialist dots inside the company on work including email marketing and customer support. When you are not working with it, a dot does what OpenAI calls proactive research, using read-only access to your connected apps, and Custom Rules let you allow an action, require approval for it or block it. OpenAI's own page adds: "always review consequential work."

Two more items from the DevDay recap are worth a marketer's attention. Business and Enterprise teams can now mention @ChatGPT in Slack and Microsoft Teams channels. And a new OpenAI Marketplace lets eligible enterprise customers put part of their existing OpenAI commitment toward approved partner software, with HubSpot and Salesforce among the first 32 partners. OpenAI puts its collective reach at 1.2 billion weekly users.

Also on September 29, OpenAI released GPT-6.1 Sol, which it says nearly matches GPT-6 Astra on agentic coding, computer use and professional work at one-fifth of Astra's standard token prices. The API price is $2 per million input tokens and $10 per million output tokens, with cached input at $0.10. Claude Sonnet 5.5, released by Anthropic on September 28, lists at the same $2 and $10 on Anthropic's pricing page. The one line where they differ is cached input: $0.10 for Sol, $0.20 for Sonnet 5.5. Sol is live in ChatGPT Work and Codex on Plus, Pro, Business, Enterprise and Edu, though not yet in regular chat.

Watch who measured what. OpenAI's page says Sol beat Opus 5.5 by 2.2 points on AutomationBench at medium reasoning effort, a test of multi-step workflows across 47 tools in sales, marketing, operations and more, at about a third of the cost, and that its share of answers with a factual error fell from 11.4% to 7.7% on deliberately hard prompts at low effort. Those are OpenAI's tests. On the independent Artificial Analysis Intelligence Index, when we checked on September 30, Sonnet 5.5 at its top setting scored 56 and GPT-6.1 Sol at its top setting scored 52. Same price, four points apart, and each company is very good at choosing the benchmark it wins.

Marketing insight: When two models cost the same, your own work breaks the tie. Run three real marketing jobs through both and let the output decide.

On the morning of DevDay, Anthropic's status page logged elevated errors across claude.ai, the desktop and mobile apps, Claude Code, Claude Cowork and the Claude API. The impact ran from 7:00 to 7:59 a.m. PT. A second issue kept many people from signing in, including through single sign-on and Sign in with Apple, and blocked new chats, file uploads and purchases. Anthropic asked anyone still signed in not to sign out, and said some messages sent during that hour may not have been saved.

Sixty minutes is not a disaster. It is, however, exactly long enough to find out which of your workflows had no plan B.

Two labs now sell their fast models at the same rate, a few points apart on an independent index, and both want those models running your work around the clock. When the engine is a commodity, the advantage moves to what you build around it: what an agent may touch, which tool takes over when one fails, and whose voice it learns. Those rules survive every model switch, and no lab can sell them to your competitor.

If you want help writing those rules, from a dot's job description to an outage plan your team will actually use, that is the exact conversation our founders like having. Talk to them below. See you in the next edition.