As OpenAI has gained extra reputation than ever because of the launch of ChatGPT final fall, its software programming interface (API) has additionally turn out to be a sought-after software for builders. Completely different corporations have discovered inspiration in OpenAI’s success and wish to replicate it, resulting in builders utilizing its APIs to construct increasingly more instruments.
To assist take care of this surge in demand, OpenAI has introduced some modifications to its API, similar to improved operate calling functionality, extra steerable variations of GPT-4 and GPT-3.5 Turbo, a brand new 16k context model of GPT-3.5 Turbo, and a 75% value discount within the Embeddings mannequin, the latter leading to lowered prices for builders that pay for the API.
Additionally: 92% of programmers are utilizing AI instruments, says GitHub developer survey
Among the new updates embody a brand new operate calling functionality within the Chat Completions API that can permit builders to attach the facility of the GPT fashions with exterior instruments extra reliably. With the replace, builders will be capable to present directions for GPT-4 and GPT-3.5 Turbo by describing the capabilities, and the mannequin will output a JSON object with the required arguments to name these capabilities.
This replace makes it simpler for builders to construct chatbots or functions that work together with exterior instruments and APIs to carry out particular duties, similar to sending emails, retrieving climate or flight data, or extracting knowledge from textual content sources similar to web sites.
Additionally: GPT-3.5 vs GPT-4: Is ChatGPT Plus price its subscription payment?
The updates additionally make for extra steerable GPT-4 and GPT-3.5 fashions, so builders can exert higher management over the fashions’ output. OpenAI is permitting builders to craft the context, specify desired formatting, and supply directions to the mannequin in regards to the desired output. Primarily, builders have extra say over the tone, type, and content material of the responses generated by the fashions used of their functions.
OpenAI additionally introduced the launch of the brand new 16k context model of GPT-3.5 Turbo, which differs from the GPT-3.5 mannequin behind ChatGPT, because it was particularly tailor-made for builders to create chat-based functions. The most recent 16k mannequin is an upgraded variant from the usual, beforehand used 4k mannequin.
Additionally: AMD unveils MI300x AI chip as ‘generative AI accelerator’
The context in “16k context” is used to explain the textual content inside a dialog that helps present context and helps the mannequin perceive the enter and supply responses which might be related to the dialog. The 4k, or 4,000, tokens in the usual GPT-3.5 Turbo mannequin restricted the mannequin’s skill to take care of the context in a dialog to a couple paragraphs. 16k, or 16,000, tokens, equals about 20 pages of textual content, giving the mannequin a bigger quantity of textual content as context.
Lastly, OpenAI introduced it has been capable of turn out to be extra environment friendly and scale back prices, so it is decreasing the Embeddings mannequin by 75% to $0.0001 per 1k tokens, and the GPT-3.5 Turbo by 25% to $0.0014 per 1k enter tokens and $0.002 per 1k output tokens. The brand new GPT-3.5 Turbo-16k mannequin is priced at $0.003 per 1k enter tokens and $0.004 per 1k output tokens.
