OpenAI wants to strengthen its role as a developer platform. The company is planning major updates that developers might like.
According to insider sources from Reuters, the upcoming updates should make it easier and cheaper to develop and sell AI-based applications using OpenAI technologies. In particular, the most powerful model, GPT-4, is expensive to run.
One innovation is reportedly the “Stateful API”, which can store the history of requests using additional memory. This would mean, for example, that entered documents would not have to be costly reprocessed for each new chat. According to Reuters sources, the cost of running an application could drop by as much as 20 times, but this will likely depend primarily on the type of application.
Similarly, OpenAI is expected to make the image analysis capabilities of GPT-4 available to developers. The Vision API will allow developers to build scalable solutions for image analysis and description. This should pave the way for other multimodal applications such as image, audio, and video processing and generation.
OpenAI will unveil its new features at its first developer conference on November 6 in San Francisco, for which OpenAI CEO Sam Altman has announced “some great things,” though not at the level of GPT-4.5 or GPT-5. OpenAI declined to comment on Reuters’ sources.
OpenAI leads commercial AI, but as a developer platform, Big Tech may be more attractive
OpenAI is by far the market leader in commercial AI chatbots with ChatGPT. But the models are expensive to develop and deploy, putting pressure on margins.
Developer services scale more efficiently, but here OpenAI is under pressure from Big Tech. Microsoft, Google, and Amazon offer access to more features and models in their cloud ecosystems, such as Meta’s Llama 2, than OpenAI, which only manages its own ecosystem.
Microsoft provides additional access to GPT models through an exclusive agreement and can provide privacy-compliant access in Europe. OpenAI must struggle to remain competitive as a model provider for developers.
The dilemma for OpenAI is that the more app developers launch their own AI apps that use models from OpenAI and others, the smaller the pie for ChatGPT and ChatGPT Enterprise becomes. At the same time, OpenAI needs to generate high revenues to fund the expensive development of new, innovative models, while also facing pressure from Google and open-source efforts.
This could be one of the reasons why OpenAI is reportedly looking into AI chip development, to reduce the cost of developing its own models and diversify its business model.