OpenAI is preparing to put its latest generation of artificial intelligence tools in the spotlight at DevDay 2026, with the company highlighting new GPT-6 models, cloud-based AI agents, realtime technology and upgraded image-generation systems ahead of the September 29 event.
According to an update sent to developers, the DevDay keynote will be livestreamed free from 10 a.m. Pacific Time on September 29, giving developers unable to attend the event in person an opportunity to follow the company’s latest announcements.
The event comes during a busy period for OpenAI’s developer platform. The company has recently announced GPT-6 Astra and is expanding its developer activities internationally, with official OpenAI developer channels listing the global DevDay programme among its September announcements.
Beyond the main event, OpenAI said it plans to take DevDay Exchange to Bengaluru, Tokyo, Seoul, Berlin, Paris, London, São Paulo and Mexico City during October and November.
The regional events are expected to bring OpenAI teams together with local developers for technical sessions, demonstrations and discussions about products being built with the company’s AI systems.
GPT-6 Astra Leads New Model Family
Among the major technologies being promoted ahead of DevDay is GPT-6 Astra, which OpenAI describes as its most intelligent model yet.
According to the company’s developer communication, Astra has been designed to improve computer use, coding and judgement, particularly for multistep work involving different applications.
The wider GPT-6 family also includes GPT-6 Sol, aimed at complex everyday tasks, computer use and coding, and GPT-6 Luna, positioned for high-volume and more clearly defined workloads.
OpenAI said API pricing for Sol and Luna is 50 per cent lower than their GPT-5.6 promotional pricing, potentially making the newer models more attractive for businesses deploying AI across high-volume applications.
The emphasis on computer use is particularly significant as the AI industry shifts from systems that mainly answer questions towards models capable of completing multistep digital tasks.
Developers are increasingly building applications where AI systems can reason over a task, use software tools, retrieve information and carry out sequences of actions rather than simply return text.
Agents API Moves Into Public Beta
OpenAI is also expanding its push into autonomous software with the Agents API, which the company said is now in public beta.
The system allows developers to build and run cloud agents using the Codex harness while OpenAI manages elements including orchestration, long-running sessions and context.
Developers can define the tools, workflows and execution environment used by their agents.
The approach is intended to reduce some of the infrastructure developers would otherwise have to build themselves when creating AI agents capable of handling longer tasks.
OpenAI cited applications such as GitHub investigations, data-warehouse queries and incident response as examples of the types of workflows developers can explore.
The development reflects a wider shift in generative AI towards so-called agentic systems, where models are expected not merely to provide information but to interact with tools and software environments to complete defined tasks.
Astra Commons Targets Developer Community
OpenAI is also expanding community experimentation around GPT-6 Astra through Astra Commons.
The initiative gives developers opportunities to participate in community build sessions, compare projects, watch demonstrations and work on applications using Astra.
A showcase gallery will allow developers to examine projects created by others and explore how different applications were built.
The community strategy could help OpenAI broaden experimentation with its newest models while giving developers practical examples beyond benchmark results and technical documentation.
GPT-Live-1 Targets More Natural Voice AI
Voice applications are another focus of the company’s latest developer push.
OpenAI’s developer update introduces GPT-Live-1, a realtime model designed to improve tool calling while producing more natural conversations.
The architecture keeps audio streaming while a backend model handles reasoning and tools, according to the material provided by OpenAI.
That separation is intended to support applications where an AI assistant needs to maintain a spoken conversation while simultaneously interacting with application data or executing tools.
OpenAI is also directing developers towards a voice-agent evaluation framework for testing spoken requests, including scenarios involving interruptions and background noise.
Such capabilities could support applications ranging from customer-service systems and digital assistants to hands-free enterprise tools.
OpenAI Expands Image Generation With Flare, Sunburst
OpenAI is also strengthening its image-generation offering with GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst.
The company says Flare is designed for faster everyday image generation and delivers higher-quality images than GPT-Image-2 while cutting latency by 50 per cent. Sunburst, meanwhile, targets more demanding creative workflows where precise editing and control are more important.
OpenAI’s current API documentation confirms that both models can generate and edit images from text and image inputs. Sunburst is positioned for workflows where editing precision is the priority, while Flare is optimised for speed and high-quality everyday generation.
The models also support a broader range of quality settings and custom image dimensions, while OpenAI’s prompting documentation emphasises improved subject preservation and targeted editing.
The image upgrades are relevant for developers building products around advertising, design, e-commerce, social content and visual editing, where maintaining consistency while making targeted changes can be as important as generating an entirely new image.
OpenAI Takes DevDay Global
The expansion of DevDay beyond a single flagship gathering also points to OpenAI’s growing focus on the international developer community.
Through DevDay Exchange, developers in eight major technology centres across Asia, Europe and Latin America will be able to participate in local technical sessions and demonstrations during October and November.
The company has not included an African city among the locations listed in the developer communication provided for this story.
For developers globally, however, the September 29 keynote livestream will provide the first major opportunity to see how OpenAI intends to position its latest tools as competition across the AI industry increasingly moves towards coding, agents, computer use, voice interaction and multimodal applications.
The combination of GPT-6 Astra, Sol and Luna, the Agents API, GPT-Live-1 and GPT-Image-2.5 suggests that OpenAI’s developer strategy is increasingly centred on AI systems capable of acting across software environments, rather than models designed primarily for standalone text generation.
That direction is likely to be one of the central themes when developers tune in for DevDay on September 29.
