The company announces ChatGPT updates to improve the bot's productivity.
OpenAI, at the company's first-ever OpenAI DevDay developer conference, announced the introduction of significant improvements and price reductions for many components of its platform. Among the new features:
Since the first release of GPT-4 in March and the opening of access to all developers in July, today starts a preview of the new generation of the model-GPT-4 Turbo. The model is trained on information about world events until April 2023 and can work with text equivalent to more than 300 pages. Performance optimization has reduced the cost of using the model by 3 times for incoming tokens and 2 times for outgoing ones compared to GPT-4.
In addition, there are improvements in calling functions: now the model is able to correctly select function parameters for performing tasks. Also, the ability to execute instructions has been improved and a new JSON mode has been added, which guarantees model responses in the form of a JSON object. Developers are now able to use the new seed parameter for reproducibility of results, which will make debugging and writing tests easier.
Other innovations include the Assistants API for creating AI-based applications and tools such as Code Interpreter and Retrieval, which make it easier to integrate knowledge and complete tasks.
New multi-modal capabilities are also introduced, including GPT-4 Turbo with image processing capability and DALL·E 3 integration for creating images via the API. In addition, developers can now generate human speech from text using the text - to-speech API.
OpenAI also offers experimental access to fine-tuning GPT-4 and launches a program for creating and sharing custom chatbots, which will allow organizations and ordinary users to train GPT-4 for their specific tasks.
Reducing prices and increasing request limits is another significant change for the company. The cost of using GPT-4 Turbo is reduced several times compared to previous tariffs. In addition, OpenAI introduces a Copyright Shield to protect against potential copyright infringement when generating content.
Such large-scale updates and lower prices will certainly increase the availability of artificial intelligence technology for a wide range of users and open up new opportunities for businesses and developers.
OpenAI, at the company's first-ever OpenAI DevDay developer conference, announced the introduction of significant improvements and price reductions for many components of its platform. Among the new features:
- The new GPT-4 Turbo model, which has an expanded context window of 128 thousand characters (equivalent to more than 300 pages), increased capabilities and lower cost;
- Assistants API for simplifying the development of native applications using artificial intelligence;
- Multi-modal platform capabilities, including image processing, image creation (DALL·E 3), and text-to-speech (TTS).
Since the first release of GPT-4 in March and the opening of access to all developers in July, today starts a preview of the new generation of the model-GPT-4 Turbo. The model is trained on information about world events until April 2023 and can work with text equivalent to more than 300 pages. Performance optimization has reduced the cost of using the model by 3 times for incoming tokens and 2 times for outgoing ones compared to GPT-4.
In addition, there are improvements in calling functions: now the model is able to correctly select function parameters for performing tasks. Also, the ability to execute instructions has been improved and a new JSON mode has been added, which guarantees model responses in the form of a JSON object. Developers are now able to use the new seed parameter for reproducibility of results, which will make debugging and writing tests easier.
Other innovations include the Assistants API for creating AI-based applications and tools such as Code Interpreter and Retrieval, which make it easier to integrate knowledge and complete tasks.
New multi-modal capabilities are also introduced, including GPT-4 Turbo with image processing capability and DALL·E 3 integration for creating images via the API. In addition, developers can now generate human speech from text using the text - to-speech API.
OpenAI also offers experimental access to fine-tuning GPT-4 and launches a program for creating and sharing custom chatbots, which will allow organizations and ordinary users to train GPT-4 for their specific tasks.
Reducing prices and increasing request limits is another significant change for the company. The cost of using GPT-4 Turbo is reduced several times compared to previous tariffs. In addition, OpenAI introduces a Copyright Shield to protect against potential copyright infringement when generating content.
Such large-scale updates and lower prices will certainly increase the availability of artificial intelligence technology for a wide range of users and open up new opportunities for businesses and developers.
