OpenAI has launched the full version of its o1 reasoning model, now available exclusively to top-tier developers in its “Tier 5” category. This access is limited to developers who have maintained an OpenAI account for at least a month and spend over $1,000 per month on the company’s services. Prior to this, developers only had access to a less powerful preview version of the model.
The full o1 model comes with high costs due to its heavy computational demands. OpenAI charges $15 for analyzing 750,000 words and $60 for generating the same amount, making it significantly more expensive than the widely used GPT-4o model. Despite the high price, the new version offers improvements, such as greater customization and a new parameter called “reasoning_effort,” allowing users to control how long the model spends on a query.
Additionally, the full model now supports function calling, linking to external data sources, as well as developer messages and image analysis, features not available in the preview. OpenAI also integrated its GPT-4o and 4o-mini models into the Realtime API, designed for low-latency speech AI applications. This API now supports WebRTC, enabling voice AI capabilities in web browsers with improved performance under varying network conditions.
Finally, OpenAI introduced direct preference optimization for developers, allowing them to refine models by submitting two responses and indicating the preferred one. This method helps models better understand user preferences, adapting to changes in formatting, style, or content depth.
These updates reflect OpenAI’s continued effort to enhance its AI capabilities, although access remains limited to its highest-tier developers.
Related topics:
Mark Zuckerberg Urges US Government to Halt OpenAI’s Shift to For-Profit Model
OpenAI Expands ChatGPT Search Feature, Posing Challenge to Google
OpenAI Responds to Musk’s Lawsuit, Claims He Pushed for For-Profit Model