Last week, reports suggested that Sam Altman-led OpenAI is going to announce several key upgrades to its large language model GPT, along with reduced prices for developers. On November 6, 2023, the company announced similar improvements, including a new GPT-4 Turbo model, which has a vast database, new APIs for enabling features like image analysis and text-to-speech conversations, and decreased prices for developers who are willing to create an app or service around OpenAI’s language model.
What’s New In GPT-4?
First, GPT-4 Turbo can now answer queries with updated information up to April 2023. In comparison to previous versions that were limited to September 2021, this is a significant upgrade, given that a lot has happened in the world since then. This will add context to the language model’s results and help users get more recent updates. Further, having a more extensive input helps keep the language model on track, so users should expect better outputs as well.
OpenAI Has Increased The Input Size To 100,000 Words
The new model also accepts much more input, up to 300 pages (up to 100,000 words), while the earlier versions only accepted up to 3000 words at a time. This allows professional users or enterprise customers to process much more data simultaneously. For example, one can ask the new model to bring back critical points from a research paper, summarize an entire book, or go through extensive data entries in one go.
Other New GPT-4 Features
GPT-4 Turbo can also accept images as inputs and enable users to generate captions, analyze the images, and read documents with stats. Developers can use this API to create apps for those with sight-related disorders. Further, one can integrate DALL-E 3 API in their product to allow users to generate images with text prompts. There’s a text-to-speech API as well, enabling developers to generate human-quality speech.
Reduced Costs To Encourage Developers
As reported earlier, Open AI has also reduced the prices for developers. The latest GPT-4 Turbo 128K model costs $0.01 per 1,000 input tokens and $0.03 per 1,000 output tokens. For those catching up, 1,000 tokens roughly equals 750 words. Further, GPT-3.5 Turbo 16K costs have decreased to $0.001 per 1,000 input tokens and $0.002 per 1,000 output tokens. Although the training price for GPT 3.5 Turbo 4K and 16K remains the same at $0.008, the input and output costs have been reduced to a fourth of what they were.
OpenAI To Provide Legal Assistance In Case Of Copyright Infringement
At the developer conference, OpenAI also announced GPTs, which enable developers to create custom versions of ChatGPT for their product, service, or app. The company also released Copyright Shield, which will step in and defend its customers if they face legal claims around copyright infringement. Moreover, GPT-4 offers a lot of upgrades for individual users and developers. It is currently in preview for developers.
You can follow Smartprix on Twitter, Facebook, Instagram, and Google News. Visit smartprix.com for the most recent news, reviews, and tech guides.
