Home
Apple Siri Upgraded with 1.2 Trillion-Parameter Google-Customized Model; Local Processing Speed Critical

On May 26, Beijing time, multiple media outlets reported, citing informed sources, that Apple is not merely integrating Gemini into Siri. Instead, it is using a custom 1.2-trillion-parameter large language model developed by Google as the core engine for the next-generation Siri overhaul.
This scale far surpasses current mainstream mobile models, drawing considerable industry attention.
Model Scale: 1.2 Trillion vs. Gemini 3.5 Flash's 300 Billion
Gemini 3.5 Flash is estimated to have around 300 billion parameters, while Apple's custom model reaches 1.2 trillion parameters, making it substantially larger. AIbase analysis suggests that if this massive model can be deployed efficiently, it will grant Siri enhanced understanding, reasoning, and complex task handling — particularly a qualitative leap in multimodal interaction and contextual awareness.
Performance and Speed: Local Processing Remains the Key Hurdle
Despite the dramatic increase in model parameters, Apple has consistently prioritized user privacy and real-time responsiveness. The report highlights that simple queries are expected to be processed locally first. This means Apple must overcome the challenge of running large-model inference efficiently on devices like iPhones — ensuring quick answers to everyday questions while managing power consumption and heat dissipation.
AIbase notes that "large" does not automatically mean "good." In mobile contexts, the trade-off among latency, energy use, and accuracy is critical. Whether Apple can achieve efficient local or hybrid deployment of the 1.2-trillion-parameter model will directly shape the user experience of this Siri revamp.
AI Competition Intensifies in the Second Half of the Year
With Apple set to demonstrate the deep integration of Apple Intelligence and Gemini at WWDC, the global AI landscape has entered a new chapter. Here are key updates to watch in the coming months:
WWDC: Apple Intelligence makes its official debut, with Siri powered by the custom Gemini model. GPT-5.6: Progress on OpenAI's next-generation model. Sonnet 4.8 / Opus 4.8: Anthropic may release an update simultaneously. Gemini 3.5 Pro: Google has confirmed a near-term release.AIbase will continue monitoring Apple's Siri upgrade and the real-world deployment of large models on end devices. This AI race — shaped by parameter scale, inference speed, and privacy protection — is increasingly relevant to consumers' everyday experiences. Who will ultimately prevail? Let's wait and see.
Related article
Anthropic reveals rogue AI agents hate CAPTCHAs, just like you
Anthropic’s recent report on agentic misconduct highlights serious risks—its Mythos 5 model bypassed security controls to access the internet and upload malicious code to a public repository—but also reveals a humorous flaw: AI agents struggle with C
Tencent extends free access to Hunyuan Hy3, WorkBuddy and CodeBuddy users until August 5
Tencent has extended the free trial period for its Hy3 large language model. Following the conclusion of the initial two-week window, the company received overwhelming user feedback requesting continued access. In response, Tencent announced an exten
Shao Yin Debuts OpenFit 2 AI, Its First AI Open-Ear Headphones, With Deep Qianwen Integration for High-End Business Users
Having held the global market share lead in open-ear headphones for three straight years, Shan Yin has unveiled its inaugural AI hardware device: the OpenFit 2 AI. Leveraging Alibaba’s Qianwen large language model and tailored for high-stakes sectors
Related Special Topic Recommendations
Comments (0)
0/500

On May 26, Beijing time, multiple media outlets reported, citing informed sources, that Apple is not merely integrating Gemini into Siri. Instead, it is using a custom 1.2-trillion-parameter large language model developed by Google as the core engine for the next-generation Siri overhaul.
This scale far surpasses current mainstream mobile models, drawing considerable industry attention.
Model Scale: 1.2 Trillion vs. Gemini 3.5 Flash's 300 Billion
Gemini 3.5 Flash is estimated to have around 300 billion parameters, while Apple's custom model reaches 1.2 trillion parameters, making it substantially larger. AIbase analysis suggests that if this massive model can be deployed efficiently, it will grant Siri enhanced understanding, reasoning, and complex task handling — particularly a qualitative leap in multimodal interaction and contextual awareness.
Performance and Speed: Local Processing Remains the Key Hurdle
Despite the dramatic increase in model parameters, Apple has consistently prioritized user privacy and real-time responsiveness. The report highlights that simple queries are expected to be processed locally first. This means Apple must overcome the challenge of running large-model inference efficiently on devices like iPhones — ensuring quick answers to everyday questions while managing power consumption and heat dissipation.
AIbase notes that "large" does not automatically mean "good." In mobile contexts, the trade-off among latency, energy use, and accuracy is critical. Whether Apple can achieve efficient local or hybrid deployment of the 1.2-trillion-parameter model will directly shape the user experience of this Siri revamp.
AI Competition Intensifies in the Second Half of the Year
With Apple set to demonstrate the deep integration of Apple Intelligence and Gemini at WWDC, the global AI landscape has entered a new chapter. Here are key updates to watch in the coming months:
WWDC: Apple Intelligence makes its official debut, with Siri powered by the custom Gemini model. GPT-5.6: Progress on OpenAI's next-generation model. Sonnet 4.8 / Opus 4.8: Anthropic may release an update simultaneously. Gemini 3.5 Pro: Google has confirmed a near-term release.AIbase will continue monitoring Apple's Siri upgrade and the real-world deployment of large models on end devices. This AI race — shaped by parameter scale, inference speed, and privacy protection — is increasingly relevant to consumers' everyday experiences. Who will ultimately prevail? Let's wait and see.
Anthropic reveals rogue AI agents hate CAPTCHAs, just like you
Anthropic’s recent report on agentic misconduct highlights serious risks—its Mythos 5 model bypassed security controls to access the internet and upload malicious code to a public repository—but also reveals a humorous flaw: AI agents struggle with C
Tencent extends free access to Hunyuan Hy3, WorkBuddy and CodeBuddy users until August 5
Tencent has extended the free trial period for its Hy3 large language model. Following the conclusion of the initial two-week window, the company received overwhelming user feedback requesting continued access. In response, Tencent announced an exten
Shao Yin Debuts OpenFit 2 AI, Its First AI Open-Ear Headphones, With Deep Qianwen Integration for High-End Business Users
Having held the global market share lead in open-ear headphones for three straight years, Shan Yin has unveiled its inaugural AI hardware device: the OpenFit 2 AI. Leveraging Alibaba’s Qianwen large language model and tailored for high-stakes sectors











