Over the course of 12 days of continuous releases, OpenAI showcased its latest advancements and achievements, with the ninth day’s announcement standing out as particularly valuable. It primarily featured the following key highlights:
**Highlights of the Ninth Day’s Release:**
1. The official release of the O1 API (previously in preview).
2. The introduction of the Real-Time API, also known as the advanced voice interaction API, accompanied by the release of an SDK to lower the barrier to entry.
3. The addition of a new fine-tuning method called “Preference Fine-Tuning.”
**Why the Ninth Day’s Release Matters:**
– **Structured Output:** This is fundamental for all Agent tools. On the ninth day, the structured output of O1 became remarkably stable, allowing the conversion of thought results into control commands.
– **Real-Time API:** Supporting structured output with a latency of less than 300 milliseconds and reduced costs, making it conducive to commercial applications.
– **Multi-End to Multi-End:** Inputs can be multimodal, and outputs can be in the form of text, voice, and Function Calling commands.
Below is a detailed interpretation of the ninth day’s release content:
**O1 Structured Output**
O1, as a powerful thinking tool, has seen its structured output become crucial for controlling real-world scenarios. Post the new release, O1’s output can reliably be transformed into control commands.
**Real-Time API**
The Real-Time API announced on the ninth day supports structured output and is suitable for scenarios requiring low latency. Additionally, the reduced cost of the API enables the commercialization of more scenarios.
**Official SDK**
The release of the official SDK has made it easier for more people to utilize the Real-Time API.
**Multi-End to Multi-End**
This is the core of the ninth day’s release, allowing both inputs and outputs to encompass multiple modalities, such as files, text, voice, and video.
**Preference Fine-Tuning**
The introduction of Preference Fine-Tuning allows AI models to actively avoid unwanted content when generating outputs, enhancing the stability of the results.
The following excerpt from a conversation further explains the significance of the ninth day’s release:
**Conversation Excerpt:**
– **On Structured Output:** Structured output has evolved from a niche toy into a core technology influencing the real world and the developer ecosystem.
– **AI Breakthrough in 2024:** Structured output has transformed from an unstable niche toy to a pivotal factor affecting the real world.
Here’s a brief recap of the 12-day release event:
**12-Day Release Event Highlights:**
– Day 1: Fully-powered O1, ChatGPT Pro membership, and more.
– Day 2: O1-based Reinforcement Fine-Tuning (RFT).
– Day 3: Introduction of Sora.
– Day 4: Launch of ChatGPT Canvas.
– Day 5: Comprehensive integration of GPT across Apple’s product line.
– Day 6: Real-time video calling, video understanding, and more.
– Day 7: Introduction of ChatGPT Projects.
– Day 8: Full launch and optimized experience of ChatGPT Search.
– Day 9: Updates on O1 API, Real-Time Voice API, and more.
– Day 10: ChatGPT’s 800 number, WhatsApp integration.
– Day 11: ChatGPT desktop version’s ability to read other applications.
– Day 12: Official release of OpenAI O3.
“`markdown
Over the course of 12 days of continuous releases, OpenAI showcased its latest advancements and achievements, with the ninth day’s announcement standing out as particularly valuable. It primarily featured the following key highlights:
**Highlights of the Ninth Day’s Release:**
1. The official release of the O1 API (previously in preview).
2. The introduction of the Real-Time API, also known as the advanced voice interaction API, accompanied by the release of an SDK to lower the barrier to entry.
3. The addition of a new fine-tuning method called “Preference Fine-Tuning.”
**Why the Ninth Day’s Release Matters:**
– **Structured Output:** This is fundamental for all Agent tools. On the ninth day, the structured output of O1 became remarkably stable, allowing the conversion of thought results into control commands.
– **Real-Time API:** Supporting structured output with a latency of less than 300 milliseconds and reduced costs, making it conducive to commercial applications.
– **Multi-End to Multi-End:** Inputs can be multimodal, and outputs can be in the form of text, voice, and Function Calling commands.
Below is a detailed interpretation of the ninth day’s release content:
**O1 Structured Output**
O1, as a powerful thinking tool, has seen its structured output become crucial for controlling real-world scenarios. Post the new release, O1’s output can reliably be transformed into control commands.
**Real-Time API**
The Real-Time API announced on the ninth day supports structured output and is suitable for scenarios requiring low latency. Additionally, the reduced cost of the API enables the commercialization of more scenarios.
**Official SDK**
The release of the official SDK has made it easier for more people to utilize the Real-Time API.
**Multi-End to Multi-End**
This is the core of the ninth day’s release, allowing both inputs and outputs to encompass multiple modalities, such as files, text, voice, and video.
**Preference Fine-Tuning**
The introduction of Preference Fine-Tuning allows AI models to actively avoid unwanted content when generating outputs, enhancing the stability of the results.
The following excerpt from a conversation further explains the significance of the ninth day’s release:
**Conversation Excerpt:**
– **On Structured Output:** Structured output has evolved from a niche toy into a core technology influencing the real world and the developer ecosystem.
– **AI Breakthrough in 2024:** Structured output has transformed from an unstable niche toy to a pivotal factor affecting the real world.
Here’s a brief recap of the 12-day release event:
**12-Day Release Event Highlights:**
– Day 1: Fully-powered O1, ChatGPT Pro membership, and more.
– Day 2: O1-based Reinforcement Fine-Tuning (RFT).
– Day 3: Introduction of Sora.
– Day 4: Launch of ChatGPT Canvas.
– Day 5: Comprehensive integration of GPT across Apple’s product line.
– Day 6: Real-time video calling, video understanding, and more.
– Day 7: Introduction of ChatGPT Projects.
– Day 8: Full launch and optimized experience of ChatGPT Search.
– Day 9: Updates on O1 API, Real-Time Voice API, and more.
– Day 10: ChatGPT’s 800 number, WhatsApp integration.
– Day 11: ChatGPT desktop version’s ability to read other applications.
– Day 12: Official release of OpenAI O3.
“`
