China's AI Models Lead Weekly API Calls for a 21st Straight Week as DeepSeek V4.1-Flash Tops the Chart, Up 219%

A weekly tally of large-model API traffic shows Chinese AI models leading for a 21st consecutive week, with DeepSeek V4.1-Flash in first place and growing 219% week over week. The figures were compiled by Futu NiuNiu.

Call-volume charts and benchmark charts are two different signals. Benchmarks measure performance on controlled tasks; call volume shows where developers actually route production traffic. The two often disagree, and volume tends to move more slowly and is harder to dress up.

The 219% week-over-week jump is the loudest number in the report. Growth of that size in a single week usually points to a version switch, with users migrating to a newly released model rather than organic expansion alone. Single-week figures are volatile, so the direction matters more than the exact percentage.

The streak is arguably more telling than the spike. Twenty-one consecutive weeks at the front suggests a persistent position in developer workflows rather than a one-off launch bump.

Call volume has clear limits as a metric. It measures usage scale, not capability, revenue or retention, and the headline does not spell out methodology, which providers are counted or how duplicate and automated traffic is handled.

Still, the trend matters for the industry. As the gaps between leading models narrow on evaluations, weekly usage becomes the harder number to move, and sustained demand feeds back into inference capacity, pricing and infrastructure planning.

Next to watch is whether the streak continues, whether competing models close the gap, and whether the growth in traffic translates into real pressure on domestic inference capacity.