ByteDance releases a new generation of AI large models with a price reduction of over 60%

Zhitong
2025.06.11 08:40

ByteDance released the Doubao large model 1.6 at the Volcano Engine FORCE Power Conference, with the price dropping to 2.6 yuan per million tokens, a decrease of 63% from the previous 7 yuan per million tokens. The new model supports multimodal capabilities and high cost-effectiveness, with an expected daily usage of 16.4 trillion tokens by the end of May 2025. CEO Liang Rubo stated that they will continue to invest and pursue intelligent breakthroughs to serve industrial applications

According to Zhitong Finance APP, on June 11, TikTok's parent company ByteDance's cloud service platform Volcano Engine held the Volcano Engine FORCE Power Conference, where the company's president Tan Dai officially released the Doubao Large Model 1.6 and announced the adoption of a unified pricing model. In terms of pricing, the Doubao Large Model has been further reduced, with a minimum price of 2.6 yuan per million tokens, a significant decrease of 63% compared to the 7 yuan per million tokens for Doubao Large Model 1.5 and DeepSeek-R1.

At the FORCE Power Conference, Volcano Engine officially released Doubao Large Model 1.6, Doubao Video Generation Model Seedance 1.0 pro, Doubao Voice Podcast Model, and Doubao Real-time Voice Model fully online, making the Doubao Large Model family a leading model with full modalities, full sizes, and high cost performance. As of the end of May 2025, the daily token usage of Doubao Large Model exceeded 16.4 trillion, an increase of 137 times compared to when it was first released in May last year.

ByteDance CEO Liang Rubo stated that developing Volcano Engine is crucial for ByteDance to become an excellent technology company and maintain technological competitiveness. In the future, ByteDance will firmly invest in the long term, pursue intelligent breakthroughs, and serve industrial applications.

In terms of pricing, Doubao Large Model 1.6 adopts a unified pricing model, pioneering pricing based on "input length" intervals. For the most commonly used input range of 0-32K in enterprises, the price for Doubao Large Model 1.6 is 0.8 yuan per million tokens for input and 8 yuan per million tokens for output, with a comprehensive cost reduction of 63% compared to Doubao 1.5 Deep Thinking Model and DeepSeek R1. The Seedance 1.0 pro model is priced at 0.015 yuan per thousand tokens, which means generating a 5-second 1080P video only costs 3.67 yuan, the lowest in the industry.

Currently, Agents that can think independently, plan, and complete tasks are gradually entering enterprise production scenarios. This not only requires models to achieve strong reasoning, multi-modality, and low cost but also necessitates upgrading AI cloud-native full-stack services. At the conference, ByteDance's Vice President of Technology Hong Dingkun and Volcano Engine President Tan Dai released 12 tools aimed at Agent development and application. Tan Dai stated that in the AI era, development paradigms, big data, computing, and security are undergoing significant changes, and the newly formed "AI cloud-native" technology stack is key for enterprises to build effective Agents and innovate development