The ninth day of holiday-themed product announcements is called "12 Days of OpenAI." OpenAI is releasing its flagship model, o1, to third-party developers through its application programming interface (API).

This is a big step forward for developers looking to create new advanced AI applications or integrate the most advanced OpenAI technology into their existing applications and workflows, whether they are enterprise or consumer.

If you're not familiar with OpenAI's o1 series, here's a quick rundown: Published again in September 2024, ChatGPT is the first in a new "family" of models from the company, offering "thinking" capabilities that go beyond the GPT-family series of large language models (LLMs).

Basically, the o1 family of models - o1 and o1 mini - take a long time to respond to user suggestions, but check themselves. as they compose the answer to avoid their correct or hallucination. At the time, OpenAI said o1 could handle more complex, PhD-level problems by real world users as well.

While developers previously had access to a preview version of o1, on top of which they could build their own apps—such as a PhD advisor or lab assistant—the full o1 production-ready release via the API improved performance and reduced runtime. , and new features that facilitate integration into real-world applications.

OpenAI About two and a half weeks ago, ChatGPT made o1 available to consumers through its Plus and Pro plansand added the ability for models to analyze and respond to images and files uploaded by users.

Along with today's launch, OpenAI announced significant updates to the Realtime API, as well as a price drop and a new refinement approach that gives developers more control over their models.

The full o1 model is now available to developers via the OpenAI API

Available as o1-2024-12-17, the new o1 model is optimized for complex, multilevel reasoning tasks. Compared to the previous o1-preview version, this release improves accuracy, performance and flexibility.

OpenAI represents significant advances on a number of metrics, including coding, math, and visual reasoning tasks.

For example, coding scores on SWE-bench Verified increased from 41.3 to 48.9, while scores on the math-focused AIME test jumped from 42 to 79.2. These enhancements make o1 suitable for building tools that streamline customer support, optimize logistics, or solve complex analytical problems.

Several new features enhance o1 functionality for developers. Structured outputs respond reliably to custom formats such as JSON schemas, ensuring consistency when interacting with external systems. Function calls simplify the process of connecting o1 to APIs and databases. And the ability to think over visual inputs opens up use cases in manufacturing, science, and coding.

Developers can also improve o1's behavior by using a new reasoning_effort parameter that controls how much time the model spends on a task to balance performance and response time.

OpenAI's Realtime API powers intelligent, conversational voice/audio AI assistants

OpenAI also announced Realtime API updates designed to enable low-expectation, natural conversational experiences such as voice assistants, live translation tools or virtual tutors.

New WebRTC integration simplifies building voice-based applications with direct support for audio streaming, noise suppression, and congestion control. Developers can now integrate real-time capabilities with minimal installation, even in variable network conditions.

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

Cached audio input costs have been reduced by 87.5%, now costing $2.50 per million input tokens. To further improve accessibility, OpenAI is adding the GPT-4o mini, a smaller, cost-effective model priced at $10 per million input tokens and $20 per million output tokens.

Text token rates for GPT-4o mini are also quite low, starting at $0.60 for input tokens and $2.40 for output tokens.

In addition to scoring, OpenAI gives developers more control over responses in the Realtime API. Features such as concurrent out-of-range responses allow background tasks such as content moderation to seamlessly run the user experience. Developers can also customize input contexts to focus on specific parts of a conversation and control when voice responses are triggered for accurate and seamless interactions.

Refinement of priority offers new customization options

Another key addition is a better settinga way to customize models based on user and developer preferences.

Unlike supervised fine-tuning, which relies on exact input-output pairs, preference fine-tuning uses pairwise comparisons to teach the model which responses are preferred. This technique is especially effective for subjective tasks, such as summarizing, creative writing, or scripts where tone and style are important.

Early testing with partners like Rogo AI, which creates assistants for financial analysts, shows promising results. According to Rogo, the priority refinements helped their model solve complex, undistributed queries better than traditional refinements, increasing task accuracy by 5%. This feature is now available for gpt-4o-2024-08-06 and gpt-4o-mini-2024-07-18, with plans to expand support to new models early next year.

New SDKs for Go and Java developers

To streamline 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, making it easier for developers to interact with OpenAI models in more programming environments. The Go SDK is particularly useful for building scalable server systems, while the Java SDK is designed for enterprise-level applications that rely on strong typing and robust ecosystems.

With these updates, OpenAI offers developers expanded tools to build advanced, custom AI-powered applications. Whether through o1's enhanced inference capabilities, Realtime API enhancements, or refinement options, OpenAI's latest offerings aim to deliver improved productivity and cost-effectiveness for businesses pushing the boundaries of AI integration.