GLM 5.2 just landed on OpenRouter.

One API call away from the best open source model out of China right now.

$1.40 input, $4.40 output per million tokens. 1 million token context window.

That pricing is a fraction of what closed frontier models charge, with a context window that actually matches them.

The catch showed up fast.

Speeds have dropped hard since launch as demand spiked. Nowhere near the roughly 300 tokens per second it shipped with.

Same story every time an open model goes viral.

Great price, great specs, then the infrastructure has to catch up to the hype.