Five Chinese AI Labs Cut Token Prices Up to 99%
Key insights
- ByteDance, Tencent, MiniMax, Alibaba, and Xiaomi all cut AI token prices between 50% and 99% within the same competitive window.
- Bank of America Securities analysts attribute the pricing race to narrowing capability differences between China's major AI models.
- Alibaba's 50% discount on Qwen3.7-Max was tied to the 618 shopping event, blending AI competition with consumer promotions.
Why this matters
When capabilities converge across all major providers, price becomes the only remaining differentiator, compressing margins and pushing AI model APIs toward commodity status. For enterprises and developers evaluating Chinese AI APIs, the 50-99% price drops across ByteDance, Tencent, MiniMax, Alibaba, and Xiaomi mean dramatically lower inference costs, potentially shifting build-versus-buy calculations in favor of API consumption over self-hosted models. Bank of America Securities analysts' framing of 'limited capability gaps across incumbents' is the key signal: China's AI market has moved from a differentiated-product competition to a pure cost race, with downstream implications for which labs survive long-term.
Summary
China's AI model market is in a full-scale pricing spiral, with five of the country's largest labs cutting token costs between 50% and 99%.
ByteD ance launched Seedance 2.0 Mini at 23 yuan ($3.40) per million tokens, half the standard rate. Tencent slashed Hy-MT2-Pro by nearly 70%, MiniMax halved its M3 series, Alibaba cut Qwen3.7-Max 50% for the 618 sales event, and Xiaomi dropped MiMo V2.5 by 99%.
Essentially: (ByteDance, Tencent, MiniMax, Alibaba, Xiaomi) are competing on price as model capabilities converge.
- Bank of America Securities sees "limited capability gaps across incumbents," making price the primary lever.
- Analysts expect declining service costs to sustain downward pricing pressure.
Capability convergence is turning China's AI model APIs into a commodity layer.
Potential risks and opportunities
Risks
- MiniMax and smaller labs without ByteDance or Alibaba's cash reserves face existential margin pressure as token prices fall 50-99% across the sector
- If Bank of America Securities is right that capability gaps are narrowing, Chinese AI labs lose the ability to justify any premium pricing, locking all players into a sustained commodity race with no exit
- Developers building on Hy-MT2-Pro or M3 series APIs face pricing uncertainty if Tencent or MiniMax revise rates upward once market share targets are reached
Opportunities
- Chinese AI application developers building on ByteDance's Seedance 2.0 Mini or MiniMax M3 APIs can sharply reduce inference costs and improve product margins immediately
- Bank of America Securities' confirmation of 'limited capability gaps' may attract enterprise buyers who previously avoided Chinese AI due to perceived quality uncertainty
- Alibaba's 618-linked Qwen3.7-Max discounts give domestic Chinese enterprises a time-bounded window to pilot and migrate AI workloads at historically low token costs
What we don't know yet
- Whether ByteDance's Seedance 2.0 Mini price of 23 yuan per million tokens is sustainable or a below-cost acquisition play to capture market share
- Which labs have the cash reserves to sustain 50-99% price cuts long-term, given that cost structures for Chinese AI inference are not publicly disclosed
- Whether Alibaba's Qwen3.7-Max discount tied to the 618 event is a temporary promotion or signals a permanent new price floor
Originally reported by scmp.com
Read the original article →Original headline: China's AI Price War Deepens as ByteDance, Tencent, MiniMax, and Alibaba Slash Token Costs Amid Narrowing Capability Gaps