OpenAI opens the most powerful features o1 to third-party developers

On the ninth day of a public holiday known as the “12 Days of OpenAI,” OpenAI is releasing its flagship version, o1, to third-party developers through its application programming interface (API).

This represents a major step forward for devs who want to develop advanced AI applications or integrate OpenAI’s advanced technology into their existing applications, whether they are business or consumer.

If you are not yet familiar with OpenAI’s o1 series, here is the list: They It was also announced in September 2024, the first in a new “family” of models from the ChatGPT company, moving beyond the main linguistic models (LLMs) of the GPT-family series and providing “conversation” skills.

Basically, the o1 family of models – o1 and o1 mini – take a long time to respond to the user’s questions and answers, but try it yourself. while making a solution to check if they are correct and avoid visual effects. At that time, OpenAI said that o1 could solve many problems, PhD problems – something provided by real world users, too.

Although developers previously had the opportunity to preview the o1 interface at the top where they could develop their applications – say, a PhD advisor or a lab assistant – the ready release of the entire o1 version through the API brings functionality, slow down. , and innovations that make it easier to integrate with real-time applications.

OpenAI had already offered o1 to consumers through ChatGPT Plus and Pro plans about two and a half weeks ago.and add the ability for models to analyze and respond to images and files uploaded by users, too.

Along with today’s launch, OpenAI announced major changes to its Realtime API, as well as cost reductions and a new optimization process that gives developers greater control over their models.

The full version of o1 is now available to developers via OpenAI’s API

The new version of o1, available as o1-2024-12-17, is designed to excel in complex, multi-thinking applications. Compared to the previous version of o1-preview, this release makes it more accurate, efficient, and flexible.

OpenAI shows significant gains in a variety of benchmarks, including text, math, and reasoning tasks.

For example, code scores on the SWE-bench Verified increased from 41.3 to 48.9, while performance on the math-oriented AIME test jumped from 42 to 79.2. These changes make o1 suitable for building tools that support customers, optimize management, or solve analytical problems.

Several new features enhance o1’s functionality for developers. Output Layers allow solutions to reliably conform to standard JSON schemas, ensuring consistency when communicating with external systems. Callbacks support the process of connecting o1 to APIs and databases. And the ability to think in visual terms opens up avenues of application in design, science, and literature.

Developers can also adjust the performance of o1 using the new parameter_effort, which controls the amount of time the model spends on tasks to improve performance and response time.

OpenAI’s Realtime API is optimized for intelligent, voice/audio AI support

OpenAI also announced changes to its Realtime API, designed to enable low-cost, natural experiences such as voice assistants, live translation tools, or virtual teachers.

The new WebRTC integration supports the development of voice-based applications by providing direct support for voice reduction, noise cancellation, and congestion control. Developers can now integrate real-time capabilities with minimal implementation, even in dynamic network conditions.

OpenAI is also introducing new pricing for its Realtime API, reducing the price by 60% for GPT-4o tokens to $40 per million input tokens and $80 per million output tokens.

The input fee for stored audio is reduced by 87.5%, which is now down to $2.50 per one million tokens. In order to improve the scalability, OpenAI is adding GPT-4o mini, a small, low-cost model priced at $10 per one million input tokens and $20 per one million output tokens.

The prices of GPT-4o mini writing tokens are also very low, starting at $0.60 for input tokens and $2.40 for output tokens.

Beyond pricing, OpenAI is giving developers the power to manage solutions in a Realtime API. Things like out-of-band solutions allow background tasks, such as content management, to run without disrupting the user experience. Developers can also adjust the input to focus on certain parts of the conversation and control when voice responses are triggered for more accurate and seamless communication.

Changing preferences provides new options

Another great addition is good choicea way to customize according to the preferences of users and developers.

Unlike supervised programming, which relies on specific types of outputs, optimization uses pairwise comparisons to train response models. This technique is especially useful for abstract tasks, such as summarizing, creative writing, or situations where tone and style are important.

Early testing with partners such as Rogo AI, which builds assistants for financial professionals, is showing positive results. Rogo also reported that preference optimization enables their model to handle complex, unclassified queries better than traditional optimization, improving performance accuracy by 5%. This module is now available on gpt-4o-2024-08-06 and gpt-4o-mini-2024-07-18, with plans to expand support to new versions early next year.

The new SDKs for Go and Java developers

To support integration, OpenAI is expanding its official SDK offerings with beta releases for Go and Java. These SDKs integrate with existing Python, Node.js, and .NET libraries, enabling developers to interact with OpenAI models in many programming environments. The Go SDK is particularly useful for building back-end systems, while the Java SDK is designed for businesses that rely on strong typing and robust environments.

With these updates, OpenAI is providing developers with additional tools to create high-quality, flexible AI-powered applications. Whether through o1’s innovative thinking capabilities, Realtime API enhancements, or streamlined workflows, OpenAI’s latest offerings aim to bring efficiency and cost-effectiveness to businesses pushing the boundaries of AI integration.