US tech firm OpenAI has unveiled an updated model that powers its ChatGPT generative AI program, promising more capacity and better performance.

CEO Sam Altman showcased the GPT-4 Turbo model at the company's first developer conference, held in San Francisco on Monday.

OpenAI says it can accept about 16 times as much input as an earlier version, and has been trained to acquire knowledge of world events up to April 2023.

The company says it will also allow the creation of customized versions of ChatGPT, and plans to make it possible for developers to sell them by as early as late November.

OpenAI also suggested paying the costs of defending its customers from copyright infringement lawsuits.

Generative AIs are trained on vast amounts of material pulled from the internet, and can respond to users' questions and prompts in natural language.

Google, Amazon, Meta and a new AI venture launched by Elon Musk are all competing with OpenAI.