OpenAI has officially unveiled its new AI model, now named OpenAI o1, representing a significant leap in AI capabilities. Previously known as Project Strawberry, this advanced model is designed to excel in complex reasoning tasks and has shown considerable improvements in fields such as math, coding, and science.
The OpenAI o1 model marks a substantial advancement by incorporating enhanced reasoning abilities. It is crafted to think through problems with greater depth and accuracy, making it especially effective at solving challenging multistep problems and generating code. Performance tests have shown that the model performs comparably to PhD students in benchmark tasks across disciplines like physics, chemistry, and biology.
However, this new model comes with a notable trade-off. Unlike its predecessors, o1 operates more slowly due to its design to deliberate thoroughly before responding. While this approach enhances its problem-solving capabilities, it means that for many routine tasks, the existing GPT-4o model may still be more efficient. This slower response time is a direct result of the model’s deeper analytical processes.
Currently, the OpenAI o1 model does not feature some functionalities present in the GPT-4o model, such as web browsing and file uploads. For now, it is limited to processing text-based prompts. This means that users seeking functionalities like image uploads or real-time web searches will find those capabilities unavailable with the o1 model.
The new model is available for ChatGPT Plus and Team users. The pricing structure for OpenAI o1 reflects its advanced capabilities, with the o1-preview API priced at $15 per 1 million input tokens and $60 per 1 million output tokens. In comparison, the GPT-4o model costs $5 per 1 million input tokens and $2.50 per 1 million output tokens. For users in India, this means accessing OpenAI o1 will cost Rs 1,260 per million input tokens, a significant increase from the Rs 420 charged for GPT-4o.
To offer a more cost-effective option, OpenAI has also introduced the o1-mini, which is designed to be faster and more affordable than the o1-preview model. Priced 80% lower than o1-preview, o1-mini is particularly suited for applications that require efficient reasoning without extensive world knowledge.
The OpenAI o1 model is currently accessible to ChatGPT Plus and Team users, with weekly rate limits set at 30 messages for o1-preview and 50 for o1-mini. OpenAI is working to expand these limits and introduce automatic model selection based on user prompts. Plans are in place to eventually extend access to the o1-mini model to all ChatGPT Free users, although specific timelines for this broader rollout have not been announced.
In a recent update, OpenAI CEO Sam Altman acknowledged that while the o1 model represents a substantial leap forward, it is not without its flaws. He described the model as “flawed” and shared performance graphs illustrating its capabilities in solving complex reasoning tasks. This acknowledgment underscores the model’s evolving nature and the ongoing efforts to refine its performance.
For developers and users, the introduction of OpenAI o1 and o1-mini provides versatile tools for a range of AI applications, from advanced problem-solving to cost-effective reasoning capabilities. The launch of these models marks an important step in the evolution of AI technology, offering enhanced features while presenting new considerations regarding cost and functionality.
