On Sept. 22, 2026, OpenAI published a post introducing two new models, GPT-6 Sol and GPT-6 Luna. OpenAI says the two models bring frontier intelligence to everyday work with different balances of capability and cost.

Source page not fetched (HTTP 403). Everything beyond OpenAI's one-line summary is unconfirmed, including prices, which plans get which model, and dates. What OpenAI has said is that there are two models: one leans toward capability and one toward lower cost. For an owner, the only question is whether either changes your bill or the quality of work you already give ChatGPT. Take an 18-person accounting firm with 6 staff on ChatGPT business seats, and assume $30 a seat a month, or $2,160 a year. If the cheaper model handles client emails and engagement-letter drafts as well as today's model, that bill should not move. If the stronger model sits behind a pricier tier, say $20 more per seat, upgrading all 6 costs $1,440 a year. That pays off only if it saves each person about 5 hours a year at $50 an hour. Firms that pay OpenAI per use through the API, or run software built on it, should care most. Owners who use ChatGPT a few times a week can ignore this until prices are published.

OpenAI published a recap of DevDay 2026 on Sept. 29, 2026, covering more than 20 announcements. The recap names GPT-6 Astra and updates to ChatGPT, Codex, the APIs, security and new tools for builders.

Facts are thin. The OpenAI page returned an error (HTTP 403), so this draft works from the feed excerpt only, and the editor should confirm details. The excerpt names a new model, GPT-6 Astra, plus changes to ChatGPT, Codex and the API. It gives no prices, plan changes or retirement dates, and those are what matter to a small business. Take a 16-person accounting firm with 6 staff on paid ChatGPT seats and one Zapier workflow that drafts client email replies through the OpenAI API. If the new model costs more per request, or the older model the workflow uses is scheduled for retirement, the bill or the automation changes. If not, nothing changes. Assume the workflow runs 400 times a month. A price change of a fraction of a cent per run comes to under $10 a month. A workflow that breaks could cost several hours of manual replies. Owners who only use ChatGPT in a browser can ignore this until the new model appears in their plan. Owners with API-connected automations should watch for deprecation notices.

On Sept. 29, 2026, OpenAI published a post introducing 'dots,' which it describes as proactive assistants that can keep working across complex projects and everyday tasks. OpenAI says dots are built to help users stay in control while work moves forward.

Details thin: OpenAI's announcement page could not be fetched (HTTP 403), so this draft works from the feed excerpt only. The editor should confirm pricing, plans and integrations. The pitch is an assistant that keeps going on a job instead of waiting for your next prompt. Take an 18-person HVAC shop whose office manager spends an assumed 5 hours a week chasing open quotes and unsigned maintenance agreements. At an assumed $30 an hour, that is $150 a week, or about $7,800 a year. An assistant that handled that follow-up on its own would be worth real money. That only holds if it can read and update the shop's dispatch and accounting systems, and if someone checks what it sends to customers. The excerpt does not say what dots cost, which ChatGPT plans include them, or what software they connect to. Owners whose staff already use ChatGPT daily should watch this. Everyone else can ignore it until those answers exist.