The next version of Apple’s iPhone operating system — iOS 27 — may allow users to select from third-party artificial intelligence models for tasks such as generating and editing text and images. According to Bloomberg, the change will also be visible on iPadOS 27 and macOS 27, which will officially become available in autumn alongside the launch of a new Apple product cycle.

The Apple Intelligence platform, introduced in 2024, currently offers ChatGPT as the only third-party option within features like Siri, Writing Tools, and Image Playground. With more options on the horizon, users will have greater freedom of choice — and Apple stands to gain a sharper competitive edge in the AI market.

Should users be permitted to choose their preferred model, a compatible AI app would need to be installed on their device, with the ability to adjust its settings thereafter. Bloomberg’s Mark Gurman reports that Apple is already internally testing integrations with AI models from both Google and Anthropic.

The development aligns with broader expectations around Siri, which is widely anticipated to undergo a significant overhaul — one in which Google’s models are expected to play a central role. ChatGPT has been the sole third-party AI model integrated into Apple Intelligence thus far, but that appears unlikely to remain the case for much longer.

Apple is also expected to inform users that it bears no responsibility for content generated by third-party models. The Bloomberg report notes that the company is preparing a dedicated App Store section to highlight compatible AI apps, while continuing to offer its own in-house models alongside them.

For OpenAI, the arrival of competing models represents a notable shift. The company has until now enjoyed exclusive third-party status on a platform spanning more than two billion devices worldwide.

Bloomberg further reports that users will be able to choose different voices for Siri conversations powered by external models — with queries handled by Apple’s own system potentially using one voice, while responses from third-party models such as Anthropic’s Claude could use another.