Byte released a large model, "99.3% cheaper than the industry"

This article is machine translated
Show original
Tan Dai, president of Volcano Engine, believes that reducing costs is a key factor in driving large models to quickly advance to the "value creation stage."

Written by: Li Xiaoyin

Source: Wall Street Journal

On May 15, at the 2024 Spring Volcano Engine Force Conference, ByteDance’s self-developed Bean Bag model was officially unveiled.

Image source: Securities Times

Volcano Engine is a cloud service platform under ByteDance. According to Tan Dai, president of Volcano Engine, Doubao Big Model includes Doubao Universal Model Pro, Doubao Universal Model Liti, Doubao Role Playing Model, Doubao Speech Synthesis Model, Doubao Sound Reproduction Model, Doubao Speech Recognition Model, Doubao Vinyl Model, and Doubao Function Call Model.

At present, the Doubao model can process 120 billion Tokens of text and generate 30 million images per day. It is reported that after a year of iteration and market verification, the Doubao model is becoming one of the largest models in China with the richest application scenarios.

The press conference stated that based on the Doubao big model, ByteDance has created the AI conversation assistant "Doubao", the AI application development platform "Button", the interactive entertainment application "Maobox", as well as AI creation tools such as Xinghui and Jimeng, and connected the big model to more than 50 businesses such as Tik Tok, Tomato Novel, Feishu, and Bytedance to improve efficiency and optimize product experience.

In addition, Volcano Engine also announced the establishment of a smart terminal large model alliance with OPPO, vivo, Honor, Xiaomi, Samsung, and ASUS.

ByteDance introduced that currently, applications such as OPPO Xiaobu Assistant, Honor Smart Office Intelligent Assistant, Xiaomi "Xiao Ai", and Douding AI Assistant for ASUS laptops have been connected to the Volcano Engine's big model service.

Large models lead to price war

While working hard to improve performance, large AI models are also continuously lowering their prices. Free services on the C-end and price cuts on the B-end are becoming a trend.

At present, the pricing of Doubao's large model is significantly lower than the industry price. According to reports, the pricing of Doubao's main model in the enterprise market is only 0.0008 yuan/thousand tokens, that is, 0.8 cents can process more than 1,500 Chinese characters, which is 99.3% cheaper than the industry price.

For reference, the market price of the same specification model is generally 0.12 yuan/thousand tokens, which is 150 times the price of the Doubao model.

Tan Dai said:

“Only with large usage can a good model be polished and the unit cost of model inference be greatly reduced.”
"The big model has changed from pricing in cents to pricing in cents, which will help companies accelerate business innovation at a lower cost."

According to the price announced by Volcano Engine, one yuan can buy 1.25 million tokens of the Doubao main model, which is about 2 million Chinese characters, equivalent to three volumes of "Romance of the Three Kingdoms".

Previously, big models including ChatGPT and Tongyi Qianwen also showed their price reduction "big move"——

On May 6, DeepSeek, a subsidiary of Huanfang Quantitative, released the second-generation MoE model DeepSeek-V2, claiming that its capabilities are comparable to GPT-4 and its price is 1% of it.

On May 9, Alibaba Cloud officially released Tongyi Qianwen 2.5, saying that individual users can use Tongyi for free through the APP, official website and mini-programs.

On May 13, OpenAI released GPT-4o, stating that it would be released to all paid and free users of ChatGPT, removing all other restrictions and reducing the API price by 50%.

Tan Dai believes that reducing costs is a key factor in pushing large models to the "value creation stage". When reducing the price of large models, it is not enough to just provide low-priced lightweight versions. The main models and the most advanced models must also be cheap enough to truly meet the complex business scenario needs of enterprises, fully verify the application value of large models, and thus catalyze innovations that go beyond existing products and organizational models.

ByteDance emphasized that Doubao is a large-scale model application that ByteDance has invested heavily in. Currently, Doubao App ranks first in AIGC applications in terms of downloads in Apple App Store and major Android application markets. According to Zhu Jun, ByteDance's vice president of product and strategy, more than 8 million intelligent entities have been created on Doubao, and the number of monthly active users has reached 26 million.

According to QuestMobile data, as of March 2024, Doubao had 23.282 million monthly active users, ranking first among AIGC apps.

Source
Disclaimer: The content above is only the author's opinion which does not represent any position of Followin, and is not intended as, and shall not be understood or construed as, investment advice from Followin.
Like
7
Add to Favorites
2
Comments