Shulou(Shulou.com)11/24 Report--

Thank CTOnews.com netizens alahu Shuaitian, rain and snow for the delivery of clues! CTOnews.com, June 14, OpenAI today announced major updates to its large language model API, including GPT-4 and gpt-3.5-turbo, including new function calls, lower usage costs, and 16000 token versions of the gpt-3.5-turbo model.

Large language Model (LLM) is an artificial intelligence technology that can deal with natural language. Its "context window" (context window) is equivalent to a short-term memory, which can store input or chat robot conversation. Increasing the size of context windows has become a technical competition in the language model, and Anthropic recently announced that its Claude language model can provide 75000 token context window options. In addition, OpenAI has developed a GPT-4 version of 32000 token, but it has not yet been released publicly.

OpenAI has just released a new gpt-3.5-turbo version of the 16000 context window, called "gpt-3.5-turbo-16k", which can handle input up to 16000 token in length, which means that it can handle about 20 pages of text at a time, which is a big improvement for developers who need to model processing and generate large blocks of text.

In addition to this change, OpenAI lists at least four other major new features:

Introducing function calling function into Chat Completions API

Improved versions of GPT-4 and gpt-3.5-turbo and "more controllable (more steerable)" versions

Reduce the price of "ada" embedded in the model by 75%.

Reduced the input token price of the gpt-3.5-turbo model by 25%

The function call function makes it easier for developers to build chat bots that can call external tools, convert natural language into external API calls, or make database queries. For example, it can convert the input "email Anya to see if she wants coffee next Friday" into a function call like "send_email (to: string, body: string)". In particular, this feature also makes it easier for API users to generate output in JSON format, which was previously difficult to achieve.

When it comes to "more controllable", a technical term for how to get LLM to behave the way you want, OpenAI says its new "gpt-3.5-turbo-0613" model will include the ability to "control more reliably through system messages." A system message is a special instruction input in API that tells the model how to behave, such as "you are Grimes and you only talk about milkshakes."

In addition to the functional improvements, OpenAI also provides considerable cost reductions. It is worth noting that the token price of the popular gpt-3.5-turbo model has been reduced by 25%. This means that developers can now use the model for about $0.0015 per 1000 token and about $0.002 per 1000 token, which is equivalent to about 700 pages of text per dollar. The gpt-3.5-turbo-16k model is priced at $0.003 per 1000 token and $0.004 per 1000 token.

In addition, CTOnews.com notes that OpenAI has reduced the price of its "text-embedding-ada-002" embedded model by 75 per cent. Embedded model is a technology that allows computers to understand words and concepts, which can transform natural languages into digital languages that machines can understand, which is important for tasks such as searching text and recommending related content.

As OpenAI constantly updates its model, the old model won't last forever. Today, the company also announced the start of phasing out some earlier versions of the model, including gpt-3.5-turbo-0301 and gpt-4-0314. The company says developers can continue to use these models until September 13, after which the old models will no longer be available. It is worth noting that OpenAI's GPT-4 API is still on the waiting list and is not yet fully open.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.