Doubao Slashes Prices and Expands Capabilities as ByteDance’s AI Offensive Accelerates

ByteDance’s AI powerhouse Doubao is ramping up its presence in the global AI race with a dual strategy of technical upgrades and aggressive market expansion, as revealed at the Volcano Engine Force Winter Conference on December 18. The tech giant unveiled new additions to the Doubao family, including a visual understanding model and a 3D model, while announcing plans to launch a longer-video-capable Doubao Video Generation Model 1.5 in spring 2025 and an end-to-end real-time voice model soon.

A key highlight from the conference was the steep price cut for Doubao’s visual understanding model: at just 3 li per 1000 tokens (equivalent to ~$0.00042 per 1000 tokens), it undercuts industry prices by 85%. This follows an earlier price reduction in May, where Doubao’s text model was priced at 0.8 li for processing over 1500 Chinese characters—99.3% cheaper than competitors. These moves have paid off: as of mid-December, the daily token usage of Doubao’s general model exceeded 4 trillion, a 33-fold increase from its launch seven months prior.

Doubao’s user growth has been equally impressive. According to data from Quantum Bits Think Tank, its daily new users surged from 200,000 to 900,000 between May and July 2024. By September, it became the first AI app in China to surpass 100 million users. In November, it recorded an average of 800,000 daily downloads, with nearly 9 million daily active users (DAU) and a monthly active user (MAU) count second only to ChatGPT. By the end of November, Doubao’s cumulative user base in 2024 had exceeded 160 million.

ByteDance’s “saturated offensive” strategy extends beyond pricing and user acquisition. The company has poured significant resources into marketing: AppGrowing data shows Doubao spent 400 million yuan (≈$55 million) on AI app advertising, making it one of the top spenders alongside Kimi. Moreover, ByteDance leveraged its Douyin (TikTok’s Chinese counterpart) ecosystem by blocking ads for competing AI apps, ensuring Doubao gets prime visibility on the platform.

On the technical front, Doubao’s visual understanding model stands out with enhanced content recognition, reasoning, and visual description capabilities. It allows users to input both text and image queries, simplifying interactions and boosting efficiency in scenarios like education (essay optimization, science popularization) and travel (translating foreign menus, explaining architectural backgrounds). Other upgrades include a music model that now generates 3-minute full works (up from 60-second simple structures) and a 2.1 version of its text-to-image model integrated with Jimeng AI and the Doubao app.

ByteDance’s AI ecosystem continues to expand, with over 20 AI apps launched in 2024 covering chat assistants, video tools, entertainment, and office solutions. These include Jimeng (competing with Sora), Maoxiang (a social AI app), Xinghui (photo editing), and Sponge Music (AI music generation).

Sources

• Jiemian News: Original Article
• Quantum Bits Think Tank
• AppGrowing
• QuestMobile
• DataEye Research Institute