Chinese AI models hit 30%+ on OpenRouter as cost gap widens

Chinese AI models have accounted for over 30 percent of traffic on OpenRouter every week since February 8, hitting 46 percent at times, a sharp rise from an 11 percent average last year. That shift accompanies a widening cost gap: OpenRouter employee Justin Summerville says Chinese open-source models run 60 to 90 percent cheaper than their US equivalents.

How big is the adoption shift on OpenRouter?

Chinese models now take a consistent, material share of OpenRouter usage: more than 30 percent weekly since February 8 and peaks of 46 percent, compared with an 11 percent average the previous year. The change is recent and sustained, not a one-off spike, and it shows up in platform traffic data rather than anecdote alone.

Platform-level share is one visible metric. Companies and developers moving traffic tell the rest of the story: Lindy moved all of its traffic from Anthropic's Claude to DeepSeek, and CEO Flo Crivello said the switch saves millions. That concrete customer migration aligns with the broader weekly share trend on OpenRouter.

How do costs and capabilities compare?

Chinese open-source models offer much lower running costs, while capability assessments put them several months behind top US models. Justin Summerville estimates Chinese open-source models run 60 to 90 percent cheaper on a per-token basis. At the same time Kyle Chan at the Brookings Institution puts the capability gap between Chinese and US models at six to nine months. The Center for AI Standards and Innovation published a May report concluding Chinese models trail leading US systems by about eight months, based on tasks in cybersecurity, software development, math, science, and abstract reasoning.

Those two strands explain the pattern: lower operating costs make Chinese models attractive for high-volume or cost-sensitive use cases even if they lag the leaders by a measurable development window. The Decoder notes per-token pricing from US providers keeps climbing, which increases the arithmetic advantage for cheaper alternatives.

Why it matters

Lower-cost Chinese models are reshaping where customers route API traffic, and the effect is visible in both platform metrics and vendor migrations. If cost savings of 60 to 90 percent persist, buyers with large token volumes or thin margins have a clear financial incentive to switch, as Lindy’s move illustrates. The Brookings and CAISI assessments mean the shift is not purely about price; capability still favors US leaders, but the gap is measured in months rather than years. That narrows the window in which US providers can win solely on performance while maintaining higher prices.

The competitive pressure could force changes in pricing strategies at US providers, or it could accelerate investment to pull performance further ahead. Either outcome will influence procurement decisions at startups, research teams, and enterprises weighing cost versus absolute capability.

What to watch

Track OpenRouter’s weekly share for Chinese models and any public pricing moves by US providers. Watch whether more customers mirror Lindy’s migration and whether subsequent CAISI or Brookings-style assessments shrink the stated six-to-nine-month gap. Signs that per-token pricing from US providers falls or that Chinese models close the capability gap would confirm the trend changing commercial AI routing.