Sam Altman's OpenAI unveils new and improved GPT-4 Turbo: All you need to know

OpenAI has unveiled a series of enhancements to its substantial language models, GPT-4 and GPT-3.5. These improvements encompass updated knowledge bases and a significantly expanded context window. Following the lead of Google and Microsoft, OpenAI also commits to safeguarding its customers against copyright litigation.

OpenAI introduced GPT-4 Turbo, currently accessible via an API preview, which has been trained with information up to April 2023. This marks a significant advancement from the previous version of GPT-4, released in March, which only incorporated data up to September 2021. OpenAI intends to release a production-ready Turbo model in the coming weeks, although an exact date has not been disclosed.

GPT-4 Turbo offers a more extensive perspective, featuring a 128K context window, equivalent to over 300 pages of text within a single prompt. Larger context windows enable GPT to comprehend questions more comprehensively and provide more thoughtful responses. Earlier iterations of GPT-4 had context windows of 8K and 32K.

OpenAI also highlighted the cost-efficiency of GPT-4 Turbo for developers. Input will cost a mere $0.01 per 1,000 tokens, the fundamental unit of text or code for large language models to process, compared to the $0.03 cost in GPT-4. Each output will be priced at $0.03 per 1,000 tokens. In summary, OpenAI asserts that this new version of GPT-4 is three times more cost-effective than its predecessors.

This latest iteration of GPT-4 retains its capacity to accept image prompts, accommodate text-to-speech requests, and integrate with DALL-E 3, a feature initially announced in October.

OpenAI highlights that the enhancements in GPT-4 Turbo empower users to instruct the model to undertake more complex tasks within a single prompt. Users can even specify their choice of coding language for results, such as XML or JSON.

Companies utilising the GPT-3.5 Turbo model will also benefit from improved context windows, functionality, and pricing. This version offers a default 16K context window and incorporates the same functional updates as GPT-4 Turbo. GPT-3.5 Turbo is priced at $0.01 for input and $0.002 for output.

OpenAI initially released GPT-3.5 Turbo in March, positioning it as the optimal model for non-chat purposes. Subsequently, in August, OpenAI introduced a version that can be fine-tuned.

In addition to the model improvements, OpenAI has announced its intention to provide copyright indemnity to enterprise users, following the examples set by Microsoft and Google. This initiative, known as Copyright Shield, signifies OpenAI's commitment to defending customers facing legal claims related to copyright infringement, covering generally available features of ChatGPT Enterprise and OpenAI's developer platform. “We will now step in and defend our customers, and pay the costs incurred, if you face legal claims around copyright infringement,” the company said in a statement.