Chinese AI models now account for over 30 percent of traffic on OpenRouter, hitting 46 percent at peak, driven by costs 60 to 90 percent lower than US providers like OpenAI and Anthropic.
Major companies including Lindy are shifting workloads entirely to cheaper Chinese alternatives.
However, independent assessments suggest Chinese models still trail leading US systems by six to nine months in capabilities like software development and reasoning.
What happened
Chinese AI models from companies like DeepSeek and Z.ai have captured over 30 percent of traffic on the platform OpenRouter every week since February 8, peaking at 46 percent. This marks a sharp rise from an average of 11 percent last year. According to OpenRouter employee Justin Summerville, Chinese open-source models run 60 to 90 percent cheaper than US alternatives.
Why it matters
US companies are shifting workloads to Chinese models to cut costs, as pricing from OpenAI and Anthropic continues to climb. The startup Lindy moved all its traffic from Anthropic's Claude to DeepSeek, with CEO Flo Crivello stating the switch saves millions. For businesses managing large-scale AI inference, the cost gap may create pressure to adopt cheaper Chinese alternatives.
What to watch
A skill gap persists—Kyle Chan at the Brookings Institution estimates Chinese models trail leading US models by six to nine months, an estimate aligned with a May report from the Center for AI Standards and Innovation (CAISI) that found Chinese AI models lag by about eight months. The assessment covered cybersecurity, software development, math, science, and abstract reasoning.
Ask the AI about this article →
The shift toward Chinese AI models reflects a fundamental cost dynamic reshaping AI adoption. As US providers maintain premium pricing strategies, companies face a widening gap between proprietary performance and open-source affordability. The data from OpenRouter—where Chinese models crossed the 30 percent threshold in early February and have sustained that level for months—signals sustained economic pressure rather than a passing trend.
However, the capability gap documented by independent assessments tempers the narrative. The six to nine month lag reported by the Brookings Institution and the May CAISI finding both point to meaningful differences in reasoning, development, and security tasks. This creates a two-tier market: cost-sensitive workloads migrate to Chinese models, while performance-critical applications may remain anchored to US systems. For businesses, the calculus increasingly hinges on which tasks tolerate a capability discount in exchange for material cost savings.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
Phonely Ltd. launched Alma, a large language AI model built for voice agents and trained on over 10 million re…
Aranya Inc., a startup founded last year, launched today with $11 million in funding
CBTS Technology Solutions LLC launched Forge Agents, a platform that turns a plain-language job description in…
Imec CEO Patrick Vandenameele said at SEMICON Taiwan 2026 that the Belgian research center is broadening its c…
Alphabet's AI Overviews now reach over 2.5 billion monthly users through Google Search, and its ad business ge…
Sarah O’Connor's book 'We Are Not Machines' explores how mechanization and AI have transformed the workforce…
