OpenAI announced Dots, their new always-on agents that connect to thousands of apps. Users can set boundaries on what these agents can do, and connecting your own machine is optional.
They also introduced GPT-6.1 Sol, a new model that offers similar intelligence to Astra but at a lower price. It costs $2 per million tokens for input and $10 for output, with faster generation speeds.
Ultrafast mode now provides up to 8 times faster output in Codex and the API, with prices increasing accordingly. This allows for quicker responses in applications.
Platform updates include new APIs like the Decisions API, which offers quick classification and routing, and improvements to Codex cloud environments. Plans were re-tiered, which caused some backlash among users.
OpenAI also tested models against benchmarks and found GPT-6.1 Sol performs well on various tasks, with fewer factual errors and lower hallucination rates. However, some models like GPT-6.1 Astra were scrapped due to safety concerns.
Why it matters
These updates improve the speed, safety, and cost of running AI models, helping developers and companies deploy models more effectively.
What to do
Review the new platform features and consider testing the ultrafast mode for your applications. Keep an eye on plan changes and evaluate their impact.
Source: https://www.latent.space/p/ainews-openai-devday-2026-dots-61