OpenAI has released a new o1 model of its API - "o1-2024-12-17", bringing many exciting updates. The model has achieved significant improvements in many aspects, including intelligent function calls, support for structured output in JSON format, and image analysis capabilities. In particular, the accuracy on mathematics and programming tasks has been greatly improved, reaching 96.4% and 76.6% respectively, far exceeding the previous version. In addition, the model's token usage in inference tasks has also been reduced by about 60%, reducing usage costs and improving running speed. OpenAI will gradually open this model to level 5 API users.
OpenAI recently launched a new version of its API o1 model, named "o1-2024-12-17". This version brings a number of exciting new features, including smart function calls, support for structured output in JSON format, and Image analysis capabilities.
According to OpenAI's report, the new model's performance has significantly improved on multiple tasks, especially in mathematics and programming. The newly released o1 model has an accuracy of 96.4% on mathematical tasks, which is significantly improved compared to the 85.5% of the previous version "o1-preview".
At the same time, the accuracy rate in programming tasks also increased from 52.3% to 76.6%. This series of improvements means that developers and users will be able to obtain more accurate and efficient results when using this model, especially when dealing with complex problems, the accuracy is better guaranteed.
In addition to the improvement in accuracy, the new model also reduces token usage by about 60% when handling inference tasks. This not only improves the running speed of the model, but also reduces the cost of use, making it more economical for developers to apply this model.
OpenAI plans to gradually open the new model to level 5 API users to ensure that users can experience the convenience brought by this new technology.
Audio API significantly reduced in priceIn terms of audio functions, OpenAI has also made important price adjustments. The cost of audio input tokens has been reduced by 60%. It currently only costs $40 per million input tokens and $80 for output tokens. The cost of caching audio input has dropped even more significantly, to only $2.50 per million tokens.
In addition, OpenAI also launched a more affordable option called GPT-4o mini, priced at $10 per million input tokens and $20 per million output tokens, providing more convenience for developing real-time voice applications.
New styles and tweaking optionsOpenAI also introduced a new method called "preference fine-tuning", which is different from traditional supervised fine-tuning. By learning the user's preferences for different responses, the model can perform better when dealing with subjective tasks such as creative writing and summarization. good.
In addition, OpenAI has also expanded its developer tools and launched new software development kits (SDKs) for Go and Java to further enhance the developer experience.
These new features will be gradually opened to developers, and OpenAI will first provide detailed documentation to Level 5 users to help them better understand and use these new features.
Highlight:
The accuracy of the new version of the o1 model in mathematics and programming tasks has increased to 96.4% and 76.6% respectively.
OpenAI audio features have seen significant price cuts, with audio input tokens dropping to $40 per million and output tokens to $80 per million.
The newly launched "preference fine-tuning" method and Go and Java SDK help developers improve their customization capabilities.
All in all, OpenAI's o1-2024-12-17 model and a series of price adjustments and feature updates provide developers with more powerful, cost-effective AI tools, further promoting the application and development of AI technology. Look forward to more exciting feature updates in the future.