OpenAI announced more improvements to its large language models, GPT-4 and GPT-3.5, including updated knowledge bases and a much longer context window. The company says it will also follow Google and Microsoft’s lead and begin protecting customers against copyright lawsuits.
GPT-4 Turbo, currently available via an API preview, has been trained with information dating to April 2023, the company announced Monday at its first-ever developer conference. The earlier version of GPT-4 released in March only learned from data dated up to September 2021. OpenAI plans to release a production-ready Turbo model in the next few weeks but did not give an exact date.
GPT-4 Turbo is cheaper to run for developers
GPT-4 Turbo will also “see” more data, with a 128K context window, which OpenAI says is “equivalent to more than 300 pages of text in a single prompt.” Generally, larger context windows let large language models like GPT understand more of the question and offer more thought-out responses. Previously, OpenAI released two versions of GPT-4, one with a context window of only 8K and another at 32K.
OpenAI says GPT-4 Turbo is cheaper to run for developers. Input will cost only $0.01 per 1,000 tokens — the basic unit of text or code for LLMs to read — compared to the $0.03 on GPT-4. Each output will be $0.03 per 1,000 tokens. Overall, OpenAI says the new version of GPT-4 is three times cheaper than the earlier ones.
This newest version of GPT-4 will still accept image prompts, text-to-speech requests, and integrate DALL-E 3, a feature first announced in October.
The company says the improvements to GPT-4 Turbo mean users can ask the model to perform more complex tasks in one prompt. People can even tell GPT-4 Turbo to specifically use the coding language of their choice for results, like code in XML or JSON.
Companies using the GPT-3.5 Turbo model will also see improvements in context windows, functionality, and pricing. It will have a 16K context window by default and the same function updates as GPT-4 Turbo. GPT-3.5 Turbo will cost $0.01 for input and $0.002 for output.
Beyond improvements to the flagship model, OpenAI also announced it will follow in the footsteps of Microsoft and Google and provide copyright indemnity to enterprise users through a program called Copyright Shield.
“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.
Copyright Shield will cover generally available features of ChatGPT Enterprise and OpenAI’s developer platform.