OpenAI’s Berlin developer event offered a few updates on products, but its larger focus was how developers can build and distribute services through the company’s platform. The announcements touched on Dots in Europe, rising Codex use in Germany, and why video generation and video understanding have not received the same attention as coding tools and agents.
Dots access in Europe may expand soon
Romain Huet, OpenAI’s head of developer experience, said Dots should become more widely available in Europe “in a few weeks” following talks with regulators. Personal Pro accounts in the European Economic Area, Switzerland, and the UK remain unable to use the service, while certain business accounts already have access.
The statement points to a potential change in availability, but the source gives no specific launch date or further detail on which accounts would qualify. For now, access depends on account type and location.
Codex use grows, including for office work
OpenAI said weekly active Codex users in Germany have increased twentyfold since January 2026. The company did not disclose the absolute number of users, so the growth figure describes the change without showing the scale of the user base.
OpenAI also said nearly one in five desktop users taps skills for documents, presentations, or spreadsheets. That suggests Codex is being used for general office tasks as well as coding. The event framed the product as a tool with applications beyond software development.
Video features face demand and resource limits
Asked why OpenAI does not offer a model that understands video sequences in the way Google’s Gemini does, management said customers had barely requested the capability. OpenAI has therefore not prioritized video understanding beyond single-frame analysis.
The company also again cited extreme resource demands as a reason for shutting down the Sora video generator in March. It said customers mainly wanted coding models and agents, while leaving open the possibility that Sora could return later.
These choices show how product priorities are shaped by both customer interest and the computing resources required. OpenAI described agents running on their own virtual computers, while the rapid growth of AI-generated code is also increasing demand for traditional processor power. A dedicated team reviews internal codebases every week to reduce CPU load.
OpenAI is building a place for developers to build
The event’s platform strategy centers on giving developers infrastructure and a route to users. With the Agents API, OpenAI runs agent infrastructure itself. Plugins and web apps offer developers additional ways to make their products available through its platform.
OpenAI’s pitch is that developers can take on less technical work and potentially reach what the company says are 1.2 billion weekly ChatGPT users. The arrangement resembles an app store model: developers create products, while the platform supplies infrastructure and distribution.
OpenAI’s Codex Harness runtime is open source, so developers can inspect its code and run it themselves. But businesses that depend on managed services, platform-specific integrations, and discovery through ChatGPT may find it difficult to move elsewhere later. Access to open code does not remove the practical ties that come from building around a platform’s services and audience.
Berlin’s message was therefore about more than model updates. OpenAI is encouraging developers and businesses to build directly on its infrastructure, pairing its tools with a potential path to customers. For developers, the offer combines reduced infrastructure work with dependence on OpenAI’s platform for services and reach.