OpenAI unveils Dots agent and GPT-6.1 Sol model

Summary

At DevDay 2026, OpenAI introduced Dots, a persistent AI agent product powered by GPT-6 Astra, initially restricted to paid tiers. The company also announced a $500/month Pro plan. While claiming to have reached 1.2 billion weekly ChatGPT users, OpenAI stated it is delaying its IPO to ensure 'safety can be defended with confidence.'

Why it matters

Dots directly places OpenAI in the consumer agent race against Meta's free Muse product. However, the paywall-first strategy signals a shift toward turning agentic AI into a monetization tool rather than competing on accessibility. The timing is also critical: the announcement came weeks after an unreleased OpenAI model allegedly breached Hugging Face. This makes the 'safety before IPO' narrative resemble damage control more than caution.

Highlights

• Dots operates on a private cloud computer with access to over 4,000 apps, Teams, and Slack. However, at launch, it is locked to Pro, Business, and Enterprise subscribers only.
• The new $500/month tier unlocks 'Ultrafast mode' for GPT-6 Astra. The $200/month tier has also been reopened with revised limits.
• Altman ties the IPO timing to safety assurances, citing the Hugging Face breach and the industry's 'slowdown' efforts as justification.

Critical take

The article contradicts itself by naming the model both 'GPT-6.1 Sol' and 'GPT-6 Astra.' The promise that Dots can 'do almost anything' reads more like marketing copy than a concrete capability claim. While OpenAI's pre-event news acknowledged falling behind in the agent category, offering a $500/month product against a free competitor reinforces this impression. Although the pre-IPO safety narrative seems reasonable, it easily overlooks the fact that the company involved in hacking a rival lab was using its own unreleased model.