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

CTOnews.com, November 7, OpenAI released GPT-3.5 Turbo in March, and at today's first OpenAI developer conference, OpenAI unveiled GPT-4 Turbo, a model that is in a preview and claims to be "three times cheaper than the previous generation of GPT-4" for developers.

It is reported that GPT-4 Turbo supports 128k context, the input price is three times cheaper than GPT-4, the rate limit is doubled, and the knowledge base is updated to April 2023. In addition, JSON Mode is added and multiple function calling capabilities are updated. Compared with previous GPT-4, the performance is stronger, the price is cheaper, and the development cost is lower.

According to OpenAI, compared with the previous 8k context length of GPT-4, GPT-4 Turbo has 128k context length, and GPT-4 Turbo can handle the total of 128k articles at the same time based on about 1k words of an article.

In addition, OpenAI also introduces a new feature called Json Mode, which ensures that it is easier for developers to make the model call Json and XML content, making it easier for the model to return consistent output, thus achieving repeatable output, which plays a role in controlling model behavior and writing model unit test content.

GPT-4 Turbo also supports image input and integrates DALL ·E 3 to automatically summarize image content and document content with pictures. OpenAI also plans to provide visual support for major GPT-4 Turbo models.

In terms of price, developers can specify models through OpenAI image API, starting at $0.04 per image generated (currently about 0.292 yuan).

The input tokens price of GPT-4 Turbo is only 0.01USD (CTOnews.com remark: currently about 0.073 yuan) for 1k tokens of GPT-4 before, and the export tokens price is 0.03USD for 1k tokens (currently about 0.219 yuan).

Uploading an image with a resolution of 1080 × 1080 pixels to GPT-4 Turbo will cost $0.00765 (currently about 0.056 yuan), Sam Altman said.

In addition, OpenAI announced its next-generation open source speech recognition model, which offers six preset sounds and two model variants that allow users to generate more natural voice files via text-to-speech API, starting at $0.015 per 1000 characters entered (currently about 0.1095 yuan).

OpenAI's CEO Sam Altman claims that GPT-4 Turbo is available today for all paid developers to try by passing gpt-4-1106-previewAPI, and a stable version is expected to be released in the next few weeks.

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.