One-line positioning
ByteDance's national-grade AI entry point — 300M MAU, #1 in China, free by default with dirt-cheap API, ready to use
Doubao is ByteDance's AI assistant. As of late May 2026 it had 302M monthly active users and 78M daily, the #1 AI-native app in China (Wenxin Yiyan 125M, Tongyi 68M, Kimi 52M, Tencent Hunyuan 41M). Doubao 4.0 supports 80+ programming languages, with code completion 94.2%, bug fix 89.7%, and code explanation 96.1%, and plugs into Toutiao, Douyin, and Xigua for real-time web search. Core features are free for individuals; developer API costs ¥2 per million input tokens and ¥6 per million output tokens.
Getting Started (based on community-tested notes, not first-person)
Note: compiled from official disclosures and user testing, rewritten and condensed by editors — not a claim that we tested each point ourselves.
- Use the app directly: pick mobile, web, or PC client; basic chat, writing, and search are all free, no card needed.
- Open Pro when needed: the paid Pro tier launched 2026-06-24 runs Standard ¥68/mo, Enhanced ¥200/mo, Premium ¥500/mo, focused on Agent-driven automated office work.
- Developers go Volcano Engine: Doubao API input ¥2/M, output ¥6/M (4.0); the 2M-context Pro version is ¥8 in / ¥24 out; Volcano Doubao 2.1 Pro is ¥6 in / ¥30 out, ¥1.2 cached.
Real Pain (community consensus)
Note: recurring pain points reported across the community, condensed by editors, not individually verified by us.
- Messy versioning: the outside world cites Doubao 4.0, the 2.1 series, and Seedance as separate lines; the same "Doubao" means different models in different articles, easy to confuse [verify].
- Free has a cost shadow: daily token calls exceed 180 trillion, the free tier's compute gap is huge, and future charges may widen.
- Opaque Pro tier: enterprise Agent power depends on specific configs; public docs describe "automated office" boundaries vaguely.
- Shorter context: versus Kimi's 1M context, Doubao's standard context is shorter; very long docs must be split.
- Search limited to ByteDance: real-time search covers Toutiao/Douyin/Xigua, raising questions on breadth and neutrality versus full web.
Hard Comparison (one-line verdicts)
- Qwen Office: Alibaba goes enterprise Agent with deep DingTalk; Doubao wins on 300M MAU national entry and free base, smoother on C-end.
- Kimi: Moonshot's 1M context + open K3 tech flow; Doubao wins on real-time search and ecosystem breadth, but deep-reasoning pricing is higher.
- Wenxin Yiyan / Tongyi: same first tier; Doubao's MAU is roughly 2–4× theirs, a clear entry-point edge.
Who It's For
- For: everyday users doing Q&A and writing, short-video/e-commerce workers riding ByteDance traffic, developers wanting cheap API, enterprises doing Agent office.
- Not for: those needing 1M long-context deep research (pick Kimi), wanting fully open self-hostable models, or averse to ByteDance-limited search.
Editor's Take
Note: our view from a C-end entry and developer-cost angle.
What's scary about Doubao isn't model strength but the entry point — 300M MAU means it's already most people's first AI app, and that habit-mind is harder to copy than parameters. API at ¥2/M input is genuinely sweet for small devs, nearly zero cost for prototypes. But we must be clear: much of its "strength" comes from ByteDance's content ecosystem (Toutiao/Douyin) feeding real-time search, not pure model dominance; version naming is also messy, so don't conflate Doubao 4.0 with Volcano 2.1 in a proposal. Our advice: ordinary users just use the free tier, devs prototype with it then compare prices, enterprises pilot Pro small-scale and see the automation boundary before scaling. Don't be lulled by "free" — at real volume the token bill makes you recalculate.
Ratings
National entry and ease of use: ★★★★★
Real-time search and ecosystem: ★★★★★
Developer API value: ★★★★★
Long-context / deep reasoning: ★★★★☆