It is difficult to say the winner of the Wanmo war now, but the price war has begun, which undoubtedly greatly reduces the cost of user use. Partners in the automotive field include more than 20 manufacturers such as Geely Automobile, Great Wall Motor, Jietu Automobile, Celis, and Zhiji Automobile.
Text丨Zhijia.com Wang Xin
Editor丨Langlang Mountain and Mingzhi Mountain
The volcano engine of the byte system has stirred up a fierce battle in the field of large models, marking an era from "pricing in cents" to "pricing in cents".
On May 15, Volcano Engine threw out a series of updates from the base model to the upper-level application at the annual "Prime Mover Congress".
Especially when showing the bean bag model family, the volcano engine did not even mention more parameters and performance-related information, but showed the epoch-making price - 0.00008 yuan/10000 Tokens.
With such a "floor price", you can buy 1.25 million Tokens of the main model of bean bags for one yuan, which is about 2 million Chinese characters, which is equivalent to three copies of "Romance of the Three Kingdoms".
Tan Bei, head of Volcano Engine, said: "According to the current pricing of the bean bag model in the enterprise market, the price of more than 1,500 Chinese characters can be processed at 0.8 centimeters, which is 99.3% cheaper than the industry. ”
Tan Bei introduced that after a year of iteration and market verification, the bean bag model is becoming one of the largest and most abundant application scenarios in China, with an average daily processing of 120 billion Tokens text and generating 30 million pictures.
The bean bag model, formerly known as "Skylark", is one of the first large models in China to pass the algorithm filing.
Based on the Doubao model, ByteDance has created the AI dialogue assistant "Doubao", the AI application development platform "Button", the interactive entertainment application "Cat Box", as well as AI creation tools such as Star Painting and Instant Dream, and connected the large model to more than 50 businesses such as Douyin, Tomato Novel, Feishu, and Giant Engine.
While serving the inside of Bytes, the bean bag model is also cooperating with many enterprise customers of Volcano Engine to co-create.
Tan believes that enterprises need models with different characteristics in different business scenarios to achieve the best results and cost performance. To this end, the Doubao model provides a model family with multimodal capabilities, which currently includes nine models, including the general model Pro, the general model Lite, the speech recognition model, the speech synthesis model, and the Wensheng diagram model.
The bean bag model is divided into two versions, the general model Pro and the lite, which are adapted to the needs of different users.
The powerful model synthesis capability and the optimization ability brought by fine-tuning make the Doubao Pro model widely used in many scenarios such as Q&A, summarization, creation, dialogue, and classification, helping enterprises to apply AI in an all-round way.
How powerful is the Bean Bao Pro model?
For example, in some scenarios, users will be very sensitive to latency and cost, and at this time, the general model of Lite is a better choice.
The advantage of the Lite model is that compared with Pro, the cost of 1,000 tokens will be reduced by 84%, and the latency will be reduced by more than 50%.
The intelligent driving of the car and the interaction of the intelligent cockpit are very suitable for the application scenario of the LITE model. Through Doubao Lite, car manufacturers can realize conversations with low latency, support multiple cockpit intelligent assistant scenarios such as chatting, search, entertainment, navigation, and vehicle control, and the effect is up to 50% higher than that of traditional voice processing, helping users get a new generation of intelligent driving experience and making cars "smarter".
This was also the beginning of the large-scale entry of Volcano engines into the game.
At yesterday's press conference, Volcano Engine set up two ecological alliances for mobile phones and automobiles. Partners in the automotive field include more than 20 manufacturers such as Geely Automobile, Great Wall Motor, Jietu Automobile, Celis, and Zhiji Automobile.
Liu Hang, head of digitalization of Jietu Automobile, also exchanged and shared his thoughts on the exploration of large models. He said that in the past two years, the team of Jietu Auto and Volcano Engine has carried out very in-depth cooperation on user operation, intelligent customer service, content community, and smart marketing.
And in the field of mobile phones, the Volcano engine is advancing faster. Large model services have been quietly stationed in the products of many partners, such as OPPO Xiaobu Assistant, Honor Smart Office Intelligent Assistant, Xiaomi "Xiaoai Classmate", and ASUS laptop Douding AI Assistant.
However, although the domestic large model is constantly refreshing the application "progress bar", the capabilities it has shown have only just taken the first step of AI agents.
Zhu Jun, vice president of product and strategy at Bytes, said frankly: "The capabilities of large models are currently amazing in many dimensions, but they are also flawed in many dimensions, and at the same time, they are evolving rapidly, and they will change greatly every three months and half a year. Zhu Jun said: "In the first half of last year, when the large model answered the questions, 6 out of 10 questions may be wrong, but this year, the illusion has been greatly reduced, and with the search engine to do knowledge enhancement, it has reached a usable state, of course, there is still a lot of room for improvement." ”
Nowadays, the application of domestic large-scale models and more commercialization directions are still focused on the to B-side.
At present, the business ideas of Volcano Engine in the application of large models can be divided into three parts: mobile phones and the automotive industry, and Volcano will choose to co-create with partners. This is also seen as a scenario with high certainty and a higher ceiling for the industry.
It is foreseeable that the large model will become a real assistant to the driver.
Write at the end:
Although it is a late entrant to the cloud computing market, Volcano Engine has shown great ambitions in the automotive field, and according to 36Kr, its automotive cloud aims to catch up with Tencent in 2025.
Volcano Engine has launched three automotive clouds in the automotive industry: intelligent driving cloud, intelligent cockpit cloud and intelligent marketing cloud.
Among them, the intelligent marketing cloud is the breakthrough point for Volcano Engine to enter the automobile market. As early as 2017, Volcano Engine began to cooperate with ByteDance's recommendation algorithm with mobile phone manufacturers, and began to introduce it to the automotive market in 2020, and its first automotive customer was a new energy vehicle brand of an American car company.
"The first point of entry for Volcano Engine is digital marketing." Yang Liwei, general manager of Volcano Engine Automotive Industry, said.
In the current era of intelligent cars, Volcano Engine has found that many car owners want to use mobile phone apps in the car. Through the volcano engine automotive industry line, the well-known APPs such as Douyin, Toutiao, Watermelon, and Chedi of Byte Ecology are "on the car", and intelligent cockpit solutions have been precipitated.
With Tesla's rapid progress in the field of autonomous driving, many domestic autonomous driving companies are also in increasing demand in the field of AI. Since 2020, Volcano Engine has made many applications using the Transformer algorithm. With BEV+Transformer becoming the standard in the field of autonomous driving in the industry last year, the Volcano Engine quickly caught up with other cloud vendors in the industry and was recognized by customers.
In the process of R&D of autonomous driving, the fastest efficiency improvement of large models is in the field of annotation, which can greatly improve efficiency by pre-labeling the collected data through the capabilities of large models. In the past cooperation between Volcano Engine and Momo Zhixing, the overall efficiency has been improved by 10 times through the pre-annotation of the Momo DriveGPT large model under the cloud service capabilities provided by Volcano Engine.
How to provide car companies with more efficient and lower-cost computing solutions has become the key to the success or defeat of cloud vendors.
Why has automotive cloud been in the spotlight in recent years? Because for cloud vendors, this is an opportunity to change the growth trend.
According to IDC's new "China Public Cloud Service Market (First Half of 2023) Tracker" report, the overall market size of China's public cloud services (IaaS/PaaS/SaaS) in the first half of 2023 will be 19.01 billion US dollars. Among them, the IaaS market size was US$11.29 billion, a year-on-year growth rate of 13.2%; The PaaS market size was US$3.29 billion, with a year-on-year growth rate of 26.3%. From the perspective of the IaaS + PaaS market, the first half of 2023 will increase by 15.9% year-on-year, the lowest year-on-year growth rate in the past three years, and the growth rate of the public cloud market in Chinese mainland will continue to slow down.
But Chinese automakers are embracing the public cloud. In May, IDC released the "China Automotive Cloud Market (Second Half of 2023) Tracking Report", which showed that the overall scale of China's automotive cloud market in the second half of 2023 reached 5.35 billion yuan, a year-on-year increase of 35.2%.
The market for cloud computing is being differentiated by automotive clouds, and the volcano engine is a major variable that cannot be ignored that affects the market landscape.