Alibaba Cuts Qwen Prices as Baidu Open-Sources Long Document OCR
Alibaba has sharply reduced Qwen API pricing, while Baidu released its long-document OCR engine as open source. Chinese models are now more compelling on both cost and capability.
Chinese AI APIs Were Priced Like US Alternatives
Until last month, Qwen API pricing was broadly in line with GPT-4o-class models, and many observers noted that the cost advantage was slim. Chinese models had been catching up fast in capability, but without a clear price edge, they struggled to motivate enterprise adoption.
In the OCR space, too, commercial-grade long-document processing tools remained expensive to access via API, with limited options for self-hosted deployment.
Two Open Strategies
Revisit RAG Costs with Qwen Price Cut
Alibaba cut Qwen prices on Qoder targeting US-hour traffic (SCMP). The move is seen as a strategic price shift aimed at the American market. Workloads with heavy token consumption — RAG pipelines, summarization batches — are now prime candidates for a fresh cost comparison.
Self-Host Document Processing with Baidu Unlimited OCR
Baidu released 'Unlimited OCR' as open source (GIGAZINE). Teams with heavy document loads should try Baidu's OSS release. The engine specializes in AI processing of long documents, and organizations can now deploy it on-premises or in their own cloud rather than relying on metered API access.
Chinese Models Get Serious About Cost Competitiveness
The Qwen price cut and Baidu's OSS release are not isolated moves — they represent a simultaneous push on two fronts: price and openness. US-based providers will find it harder to hold their existing pricing tiers.
For teams evaluating AI infrastructure, now is the right moment to build a proper comparison. Cost alone is not the only factor — data residency and compliance requirements must be weighed alongside price and capability.
Sources: SCMP (Qwen price cut coverage), GIGAZINE (Baidu Unlimited OCR release)