OpenAI Unveils GPT-6 Sol, Luna, and 'Dots' Agents at DevDay 2026

Overview

OpenAI has officially expanded its GPT-6 ecosystem with the release of two new models, GPT-6 Sol and GPT-6 Luna, alongside a significant leap in autonomous agent capability through its new "Dots" feature. Announced at DevDay 2026, these updates represent a strategic push to make frontier AI models faster, more affordable, and deeply integrated into daily workflows.

The company emphasized that the primary goal of this rollout is to democratize access to high-performance AI. By introducing tiered models that balance cost and speed, OpenAI aims to support both individual productivity and large-scale enterprise operations. The simultaneous launch of Dots, powered by GPT-6 Astra, marks a shift from passive chatbots to proactive, always-on digital assistants.

Pricing and Model Accessibility

The most immediate impact of this release is on pricing and accessibility. GPT-6 Sol and GPT-6 Luna are designed to bring the core capabilities of the flagship GPT-6 Astra model to a broader audience without the associated high costs. According to OpenAI, the company has optimized caching and inference efficiency, passing those savings directly to users.

• 50% price reduction: API prices for both Sol and Luna are half the promotional rate of GPT-5.6.
• Tiered access: Sol and Luna are available in ChatGPT Work and Codex for Plus, Pro, Business, and Enterprise users.
• Free tier inclusion: Free and Go users can now try GPT-6 Luna in the desktop application.

This strategy aligns with OpenAI’s stated mission to ensure that the benefits of general-purpose AI are widely shared. By lowering the barrier to entry, the company hopes to accelerate adoption across diverse sectors, from small businesses to large legal firms.

Introducing Dots: The Evolution of Autonomous Agents

Alongside the model updates, OpenAI unveiled Dots, a new class of agents built on the GPT-6 Astra foundation. Dots are described as "remarkably capable, always-on agents" that operate with the initiative of a high-agency engineer or chief of staff. Unlike traditional assistants that wait for explicit commands, Dots are designed to learn user preferences and execute tasks proactively in the background.

The architecture of a Dot includes its own isolated computer environment. This allows it to work securely across 4,000+ applications through ChatGPT plugins. From booking a dinner reservation to managing complex project pipelines, a Dot can handle a wide spectrum of responsibilities.

"Your dot can do simple things like book a table—or take on your most ambitious work with the initiative of a high-agency engineer or chief of staff."

Control and Safety Architecture

OpenAI stressed that safety and user control are central to the Dot experience. The company implemented several safeguards to prevent unauthorized actions or data leaks:

1. Isolated Cloud Computer: Each Dot runs on its own secure cloud instance, separating it from the user’s primary devices.
2. Optional Connectivity: Connecting personal apps is entirely optional. Users can start with a basic Dot and expand capabilities over time.
3. Granular Boundaries: Users can specify exactly what a Dot can do autonomously, when it must ask for permission, and what actions are strictly forbidden.
4. Contextual Memory: The agent learns from interactions to anticipate needs, reducing the friction of repetitive tasks.

Dots are rolling out globally in ChatGPT on web, mobile, and desktop platforms for eligible Pro, Business Premium, and Enterprise users.

ChatGPT Voice Gets a Major Upgrade

The DevDay announcements also included a significant update to ChatGPT Voice, enhancing its utility for hands-free productivity. Previously limited in scope, Voice now supports hundreds of plugins, including integrations with email, calendar, and Slack.

Powered by GPT-6 Astra, Sol, and Luna, the updated Voice interface is now fully integrated with ChatGPT Work. This allows users to create documents, presentations, spreadsheets, and websites using only their voice. The feature supports complex tasks performed directly in the browser, making it a viable primary interface for many professional workflows.

The rollout is available globally today via the latest version of the ChatGPT app, marking a step toward more natural, multimodal human-computer interaction.

Commitment to Research and Safety

Beyond commercial products, OpenAI continued to highlight its commitment to rigorous research and safety standards. During the event, the company demonstrated how frontier models perform on MentalHealthBench, a new open benchmark designed to evaluate AI interactions in realistic mental health conversations.

The benchmark was developed with input from more than 80 mental health clinicians. OpenAI released the tool openly to encourage the broader research community to examine methods, run independent evaluations, and build upon the findings. This move underscores the company’s stated priority of aligning AI interests with human well-being, particularly in sensitive domains where error rates can have serious consequences.

Additionally, the company showcased specific vertical applications, such as Astra for Law. This offering pairs the Astra model with a Legal Search Index covering over 230 million URLs of U.S. case law, statutes, and regulations. Paired with 26 partner-built plugins, it aims to streamline legal research and analysis, demonstrating how specialized tools can enhance professional expertise rather than replace it.

As competition in the AI sector intensifies, these updates position OpenAI to capture both the high-end enterprise market and the mass consumer segment. With GPT-6 Sol and Luna driving down costs and Dots raising the bar for automation, the gap between experimental AI and practical, everyday utility continues to narrow.