Alibaba cuts prices on its visual language model Qwen-VL by up to 85%, after cutting Qwen prices by as much as 97% in May, amid growing AI competition in China
Alibaba is cutting prices on its large language models by up to 85%, the Chinese tech giant announced Tuesday.
Context & Ripple Effects
Alibaba had already paired AI rollout with cloud-cost reductions, including planned cuts to core cloud-product costs, and later broadened Qwen availability through more than 100 open-source Qwen 2.5 models. The Qwen-VL reduction extends that playbook from access and distribution to paid model pricing.
The repeated reductions matter because they make the cost of deploying multimodal AI a more explicit competitive lever for Alibaba, rather than leaving model selection to capability alone.
First-order effects
- Qwen-VL customers face sharply lower listed model costs, reducing the immediate budget hurdle for visual-language workloads.
- Alibaba accepts lower revenue per unit of Qwen-VL usage in pursuit of greater adoption and usage volume amid Chinese AI competition.
Second-order effects
- Competing Chinese model providers face pressure to respond on pricing, model access, or bundled cloud offerings when customers can use Alibaba's reduced rates as a benchmark.
- Enterprise buyers gain leverage in model negotiations and can test more multimodal use cases at lower marginal cost.
Third-order effects
- If repeated cuts persist, model inference is likely to become more commoditized, with providers differentiating through cloud integration, distribution, and application ecosystems rather than API price alone.
- The pattern increases the importance of sustaining efficient inference economics: lower prices can expand demand, but also tighten the margin available to model and cloud providers.
The trend: China's AI market is moving toward price-led competition in which falling inference costs and broad model availability shift buyer power toward enterprises and developers.