OpenAI has announced several new additions and improvements to their platform. The highlights include the launch of GPT-4 Turbo, which is more capable and cheaper than its predecessor and supports a 128K context window. The Assistants API has also been introduced, making it easier for developers to build their own AI apps with goals and function calling capabilities. Other updates include improved instruction following, JSON mode, reproducible outputs, and log probabilities. The platform also offers the updated GPT-3.5 Turbo model and introduces new multimodal capabilities such as vision, image creation (DALL·E 3), and text-to-speech. Overall, these updates enhance the AI capabilities, customization options, and pricing structure of the OpenAI platform.
| Signal | Change | 10y horizon | Driving force |
|---|---|---|---|
| Introduction of GPT-4 Turbo | Improved and cheaper language model with larger context window | More powerful and affordable language models | Improving AI accessibility and affordability |
| Advancements in function calling | Ability to call multiple functions in a single message | More efficient and accurate function calling | Enhancing productivity and streamlining interactions with AI models |
| Improved instruction following and JSON mode | Better performance in following instructions and generating valid JSON | Enhanced accuracy and compatibility in generating specific formats | Improving the reliability and usability of AI-generated outputs |
| Assistants API and new capabilities | Enables developers to build agent-like AI experiences with specific instructions and tools | Creation of agent-like AI apps with custom functionalities | Enhancing the versatility and capabilities of AI applications |
| New modalities in the API (vision, DALL·E 3, TTS) | Integration of image processing, image generation, and text-to-speech capabilities | More immersive and diverse AI-powered experiences | Enhancing the richness and functionality of AI applications |
| Model customization options (fine-tuning and custom models) | Access to fine-tuning GPT models and custom model training | Greater adaptability and customization for specific domains | Meeting the unique needs and requirements of organizations |
| Lower prices and higher rate limits | Reductions in pricing and increased token usage limits | More affordable and scalable AI usage | Improving affordability and scalability for AI developers |
| Introduction of Copyright Shield | OpenAI backs and defends customers against legal claims around copyright infringement | Protection and support for customers in legal disputes | Providing customer support and legal safeguards |
| Release of Whisper v3 and Consistency Decoder | Improved automatic speech recognition model and advanced image decoder | Better performance and quality in ASR and image generation | Advancing speech and image processing technologies |