Hajra Naz
According to an announcement from OpenAI, all paid API users can now access the GPT-4 API. It also talked about new chat-based models, said that the Completions API would be replaced by the Chat Completions API, and said that older models would be retired.
The GPT-4 API, which was first released in March and is generally regarded as its most powerful API product, has been in closed testing up until this point. As a Programming interface, developers can utilize an exceptional connection point to coordinate OpenAI’s enormous language model (LLM) into their own items for utilizations like rundown, coding help, examination, and composition.
The model provides Internet-based output to other applications and runs remotely on OpenAI servers. OpenAI says the GPT-4 API with an 8K setting is available to existing developers who have a fruitful installment history, with plans to open admittance to new developers toward July’s end.
It has also decided to start retiring “Completions API” models in favor of newer Chat Completions API models in order to distance itself from older GPT-3-style models.
It is claimed by OpenAI that its Chat Completions API models now account for 97% of API GPT usage since its launch in March. This deprecation plan includes resigning models that are essential for the Completion API in a half year. Beginning on January 4, 2024, developers using some of these older models will have to manually upgrade their integration. These older models don’t converse in a chat-like manner; they just answer a prompt.
The older models will still be available for the time being, but they will be referred to as “legacy” products. A partial list of models that will no longer be supported includes:
- ada
- babbage
- curie
- DaVinci
- DaVinci-instruct-beta
- curie-instruct-beta
- text-ada-001
- text-babbage-001
- text-curie-001
- text-DaVinci-001
- text-DaVinci-002
- text-DaVinci-003
These more experienced models will receive drop-in replacements from OpenAI. The text-davinci-003 model should be replaced with the gpt-3.5-turbo-instruct model, according to OpenAI. Essentially, the ada-002, babbage-002, curie-002, and davinci-002 models will supplant prior renditions of each. OpenAI claims that applications that make use of some of these GPT-3 models—like ada, babbage, curie, and DaVinci—will “automatically be upgraded to the new models listed above on January 4, 2024.”
OpenAI also stated that it will make APIs for Whisper, DALL-E, and GPT-3.5 Turbo “generally available” “based on the stability and readiness of these models for production-scale use.” Additionally, the company anticipates that model refinement will continue throughout the year.
