Apple has a 2025 hatchet plan to drop their on-device LLM, however, there’s already a competitor lurking in the background – Part 1

Getting your Trinity Audio player ready...

Apple has been building up it’s on-device LLM technology for five years with Apple Silicon, with it’s NPU that was introduced in 2017. Same time, Huawei introduced their NPU chip in September 2017, before Apple, as the world’s first chip with NPU integrated. Since then both companies have incrementally been adding on-deivce NPU features into their software at a small scale over the years on their products, with Apple more pronounced on Apple operating systems like macOS, iOS etc, besides “Siri” meanwhile, Huawei has been incrementally building up it’s use of focus on camera technologies, image editing features, AI powered AR recognition features with services in the cloud, mixed processing on EMUI Android-based software over the years that has been morphed into earlier versions of rebranded custom hybrid OpenHarmony/LiteOS-based, AOSP mobile ROM of HarmonyOS 2 up to version 3.1. Once HarmonyOS 4 hit last year, it progressively went to the higher stage of introducing LLMs in the company’s virtual assistant such as Celia with chatbots interfaces.

On July 19 2023, Bloomberg reported Ajax Generative LLM technology from Apple in development. Apple’s Ajax system is built on top of Google Jax, the search giant’s machine learning framework. Apple’s system runs on Google Cloud, which the company uses to power cloud services alongside a mixture of its own infrastructure and Inc.’s AWS.

As part of its recent work, Apple is seeking to hire more experts in generative AI. On its website, it is advertising for engineers with a “robust understanding of large language models and generative AI” and promises to work on applying that technology to the way “people communicate, create, connect and consume media” on iPhones and its other devices.

An ideal spot for Apple to integrate its LLM technology would be inside Siri, allowing the voice assistant to conduct more tasks on behalf of users. Despite launching in 2011, before rival systems, Siri lagged competitors as Apple focused on other areas and adopted fewer features in favor of privacy.

On August 4, 2023 at Huawei Developer Conference Huawei beats both Google and Apple, as Huawei brings large AI model to mobile voice assistant in China, which is the Chinese version of Celia called Huawei Xiao Yi in Chinese.

On September 2023 reported by, Apple became the biggest buyer of AI companies.

On January 18, 2024, Huawei announces HarmonyOS NEXT system successor of current HarmonyOS for the next version of the operating system for the fall of 2024 at the Hongmeng Ecology Conference for developers in China, that includes a full AI system stack for on-device multimodal LLM that packs both in-house Pangu AI Model and MindSpore AI Framework with full developer SDK access with forked TypeScript, ArkTS and in-house Cangjie programming language support taking advantage of existing older and newer architectures of in-house Kirin chipsets that has Neural Processing Units, with software that includes the Neural Network Runtime system, alongside in-house Huawei Cloud stack and open-source Cloud stack by Eclipse for cloud features outside offloading processing with open-source OpenHarmony and Oniro OS variants of HarmonyOS codebase that are morphed into on-device LLM-centric operating systems. Focused in the Chinese market, post-sanctions with software while preparing to return globally in phone hardware this year with the last EMUI 14 Android software that carries Celia LLM assistant.

On March 14, 2024 Apple Buys Canadian AI Startup as It Races to Add Features according to Bloomberg.

On March 18, 2024, Apple researchers build multimodal LLM as AI strategy takes shape. The project is called MM1, Apple researchers used a mix of text-only data, images paired with captions and longer documents with relevant images to train the model. Researchers found fine-tuning the model with high-resolution photos improved outputs as well. Reported via CIO Dive.

On the same day, Apple Is in Talks to Let Google Gemini Power iPhone AI Features via Bloomberg as a trial with expected system and Siri improvements with it on iOS 18 for the bigger plan afterwards. On March 19 2024, it’s been reported that Huawei is preparing for the second stage of it’s LLM plans, for it’s first half camera photography focused flagship called P70 with HarmonyOS 4.1, that features both mixed cloud-client side LLM for various tasks of the AI system of APIs introduced on HarmonyOS 4.0 for HarmonyOS APIs in China by the middle of next month, in April 2024.

Some point out towards the year 2025 as been reported last December, Apple’s bigger plans for it’s in-house LLMs as the launch year. The same year when Huawei is expected to fully return to the global market with the newer advanced HarmonyOS NEXT iteration operating system, with a big daunting task to get developers on board outside of China secured market where Huawei serves outside North America, US and Canada, by the time Huawei breaks all sanctions, and blacklisting restrictions on American telephony app framework software, closed component of Google Mobile Services apps, after semiconductor chips.

The battle between two logger heads from Cupertino to Shenzen continues into 2025 on winning consumers in the consumer market. The battle begins between two developer conferences, at Apple’s WWDC 2024 vs Huawei’s HDC 2024 in the summer of 2024, around June.

This is part of 1 of a two part series.

Apple’s competitor multimodal LLM efforts and plans to conquer firecely competitve global markets with on-device AI stack – Part 2