
The Blue Heart Grand Model is a self-developed generic grand model matrix released by Vivo.
History and Background
- On November 1, 2023, Vivo officially released its self-developed big model matrix "Blue Heart Big Model" at the Vivo Developer Conference 2023, and at the same time released the first global intelligent assistance Blue Heart Little V developed based on the Blue Heart Big Model, the natural language dialogue robot Blue Heart Thousand Inquiries, and other brand new functional applications.
 - On December 14, 2023, Vivo officially released the Vivo S18 series of cell phones, the first batch of the series to carry the Vivo AI Blue Heart Large Model 7B.
 - On December 27, 2023, Vivo officially released the iQOO Neo9 series of cell phones, which are also the first batch to carry the self-research AI Blue Heart Grand Model 7B.
 - In February 2024, the State Internet Information Office released the fourth batch of in-depth synthesis service algorithm filing list, "vivo blue heart big model algorithm" is listed.
 - On October 10, 2024, Vivo officially released its new self-researched Blue Heart Big Model Matrix at the Vivo Developer Conference 2024, comprehensively upgrading the language big model and end-side big model capabilities and bringing Vivo's self-researched Speech Big Model, Image Big Model, and Multimodal Big Model.
 
Model Components and Features
Blueheart big model contains several self-developed big models, which comprehensively cover the core scenarios and meet different application requirements and arithmetic conditions:
- Blue Heart Large Model 1B: Billions of models.
 - Blue Heart Large Model 7B: The seven-billion-level model is a dual-purpose end-cloud model built for cell phones that performs well in scenarios such as language understanding and text creation.
 - Blue Heart Large Model 70B: A seventy billion level model that excels in role-playing, knowledge quiz, and other scenarios.
 - Blue Heart Large Model 130B: Thirteen hundred billion level modeling.
 - Blue Heart Large Model 175B: Seventeen hundred and fifty billion level models.
 
In addition, the new 3 billion Blue Heart end-side large model 3B can cross the line to match the industry 7B~9B model in terms of dialog writing, abstract summarization, information extraction and other capabilities. Compared with Blue Heart 7B, Blue Heart 3B extreme performance is improved by 300%, power consumption is optimized by 46% in balanced mode, and memory is reduced by 63%, achieving an extreme word output speed of 80 words/s, with a system power consumption of only 450mA, and a memory occupation of only 1.4GB.
Technology Principles and Architecture
- The technical principles of the Blue Heart Big Model are based on deep learning frameworks, such as TensorFlow or PyTorch, to build and train complex neural networks.
 - The model learns the patterns and structure of the language through large-scale pre-training using the Transformer architecture, a model structure that has performed well in natural language processing tasks.The attention mechanism in the Transformer model allows the model to better understand and process the relationships between different parts of the input data.
 - The Blue Heart Big Model also uses an end-to-end learning approach, where the entire process from input to output is continuous and requires no human intervention.
 
Application Scenarios and Functions
The Blue Heart Big Model and its derivative applications play an important role in a number of areas, including cell phones:
- Blue Heart Little V: As the first global intelligent assistant application of the Blue Heart Big Model, Blue Heart Little V has the ability to be the exclusive personal assistant of the cell phone. It can receive and process information data by voice, text or drag-and-drop, and it is also designed with a convenient and compact hover state, which makes it easy for users to open or minimize and hang up at any time.
 - thousands of inquiries from the blue heart (idiom); inquiring after the death of a noblemanThe natural language dialog robot independently developed by Vivo supports natural dialog with users, quick Q&A of knowledge information, literary creation, Chinese comprehension, picture generation and other functions, and can also act like a software engineer to write applications.
 - Blue Heart Voice Large Model: It can accurately understand natural semantics, simulate the human voice, and express emotions to support simultaneous interpretation.
 - Blue Heart Image Large Model: It strengthens the ability to generate a fusion of Chinese characteristics and oriental aesthetics, and is one of the most knowledgeable image models in China about the Chinese context. It supports national style ink art and can add Chinese character related creations in the image generation process.
 - Blue Heart Multimodal Large Model: Upgraded visual perception and understanding to provide a deeper understanding of the phone screen, delivering a smoother and more natural experience in real-time conversations over video streams.
 
Strengths and Achievements
- Blueheart Big Models has been ranked No. 1 in terms of overall capabilities on several big model lists such as SuperCLUE, CMMLU, C-EVAL, etc., showing its leading position in the industry.
 - Vivo's AI capabilities have covered more than 60 countries and regions around the world, serving more than 500 million cell phone users, with more than 3 trillion big model token outputs.
 
data statistics
Related Navigation

Shangtang Technology has launched a comprehensive big model system with powerful natural language processing, text-born diagrams and other multimodal capabilities, aiming to provide efficient AI solutions for enterprises.
                    
XiHu LM
Westlake HeartStar's self-developed universal big model, which integrates multimodal capabilities and possesses high IQ and EQ, has been widely used in many fields.
                    
Congrong LM
The multimodal large model independently developed by CloudScience has the ability of real-time learning, synchronous feedback, cross-modal interaction, etc. It is widely used in many industries such as finance, security, government affairs, etc., to promote the popularization and development of AI applications.
                    
s1
An AI model developed by Fei-Fei Li's team that achieves superior inference performance at a very low training cost.
                    
OpenAI o3-mini
OpenAI introduces small AI models with inference capabilities and cost-effective pricing, designed for developers and users to optimize application performance and efficiency.
                    
Hunyuan T1
Tencent's self-developed deep thinking models with fast response, ultra-long text processing and strong reasoning capabilities have been widely used in intelligent Q&A, document processing and other fields.
                    
ERNIE X1 Turbo
Baidu has launched a new generation of high-level AI assistants to disassemble complex tasks and automate the entire process with autonomous deep thinking, multimodal toolchain invocation and extreme cost advantages.
                    
TianGong LM
Kunlun World Wide's self-developed double-gigabyte large language model, with powerful text generation and comprehension capabilities and support for multimodal interaction, is an important innovation in the field of Chinese AI.
                    No comments...
