Opeli made its first devay, in which many amazing new functions were announced.
The OpenAI Devday conference hosted the director of the company General Sam Altman. He emphasized that ChatgPT became one of the fastest growing consumer products and achieved 100 million users a month a few months after its launch. Today, over 100 million people use chatgpt every week. Over 2 million programmers create their solutions using API Chatgpt and Whisper by OpenAI.
Here is a summary of OPENAI's main ads in case you miss this event.
Turbo GPT-4 start
Users will soon access the new model – GPT -4 Turbo. This model provides a 128K context window, has a lower cost and contains the equivalent of over 300 pages of the text in one line. Turbo GPT-4 includes knowledge about world events until April 2023. The prices of input and output tokens have been reduced 3 and 2 times, respectively, compared to GPT-4.
API assistants
Developers can now create specialized AI assistants who implement specific instructions, use additional knowledge from the user base and can call models and tools to perform their tasks (for example, Python code interpreter).
In fact, this is an opportunity to personalize the OpenAI model for your tasks and knowledge base.
Multimodal possibilities in API
In the Turbo GPT-4 you will be able to send not only the text, but also images via the API interface and create poems. It will also be possible to generate images using the API request and converting the text to human speech using the Speech Synthesis API interface.
Playful model calls
API calls now have a grain parameter – a beta function that allows you to get repetitive results, which is useful for reproducing demands to debug, write more varied unit tests and generally monitoring the behavior of the model.
Login probability
Within a few weeks, Opeli plans to launch a function that returns the likelihood of each token in Turbo GPT-4 and Turbo GPT-3.5. This will be useful for developing functions such as autocomplete in search engines, as well as to detect hallucinations of the model.
More tokens per minute
Limits per minute for paid GPT-4 users have been doubled. New limits are available according to a layer of use.
Custom models
A new program is available for organizations that require more personalization than the usual model configuration. The program allows you to cooperate with OPENAI researchers to create non-standard GPT-4 models for specific domains.
