OpenAI's o1 reasoning model can now be integrated with apps
What's the story
OpenAI, the leading artificial intelligence research lab, has announced the integration of its advanced "reasoning" AI model, o1, into its API.
However, the initial rollout is only available for developers in OpenAI's "tier 5" usage category.
To qualify for this tier, developers would have had to spend at least $1,000 with OpenAI and kept an account for more than 30 days since their first successful payment.
Unique features
o1: A self-checking AI model
The o1 model is unique in that it can fact-check itself, a feature that prevents it from falling into common traps of other AI models. However, the self-checking mechanism can also lead to longer processing times for solutions.
The new model replaces the previous version, called o1-preview, which was already available in the API but didn't have some advanced features of its successor.
Pricing details
High cost and customization of o1 model
The o1 model is way more expensive than other AI models owing to its high computing resource requirements.
OpenAI charges $15 for every ~750,000 words analyzed by o1 and $60 for every ~750,000 words generated by the model. This is six times more than OpenAI's latest "non-reasoning" model, GPT-4o.
Despite the cost, the o1 model comes with enhanced customization features like function calling, developer messages, image analysis capabilities, and an API parameter called "reasoning_effort."
Expansion strategy
OpenAI's future plans for o1 model
OpenAI has disclosed that the version of o1 integrated into its API and soon to be integrated into its AI chatbot platform, ChatGPT, is a "new post-trained" version.
This updated model, "o1-2024-12-17," improves on "areas of model behavior based on feedback."
The company plans to gradually extend access while working to expand access to additional usage tiers and increase rate limits.
Model improvements
Enhanced performance of o1 model
The latest version of the o1 model promises "more comprehensive and accurate responses," especially for programming and business-related questions.
It is also less prone to incorrectly refusing requests, thus improving its overall performance.
These improvements are part of OpenAI's continuous efforts to refine and improve its AI models based on user feedback and technological advancements.