Apple is ramping up its artificial intelligence efforts with a massive cloud-based language model and a new internal chatbot — but Siri still seems stuck in limbo.
According to a new report from Bloomberg, the company is currently testing multiple large language models (LLMs), including a flagship version with a staggering 150 billion parameters — nearly matching the scale of OpenAI’s ChatGPT.
Unlike Apple Intelligence, which relies on smaller, on-device 3 billion parameter models, this new LLM runs in the cloud and is designed for more powerful reasoning and conversational abilities. Early testing via Apple’s internal Playground tool shows it performs comparably to recent versions of ChatGPT, though issues like hallucinations (false or misleading responses) persist.
Apple is also developing a new AI chatbot codenamed “Knowledge“, which pulls real-time information from the internet. This tool is being led by Robby Walker, the former head of Siri. However, like many AI projects at Apple, it too faces technical and architectural hurdles that prevent it from replacing the legacy Siri system — at least for now.
Sources say Apple teams in Switzerland are working on an entirely new architecture for Siri, aiming to finally move beyond its outdated backend. Still, it’s unclear when — or if — these tools will be revealed, and they may not even be mentioned at the upcoming WWDC.
Meanwhile, to expand its AI footprint, Apple is preparing to integrate Google’s Gemini model into iOS 26, while also reportedly in talks with Perplexity AI for a future collaboration.
With AI becoming the next battlefield in tech, Apple is clearly investing heavily — but it still has ground to cover if it wants to lead the conversation.