Skip to content

openrouter: glm-5.3 flash volume halved, still #1

Illustrated night race: neon motorcycles labeled GLM, LUNA, DEEPSEEK, HY4 and HY3 battling for the OpenRouter token lead. models

openrouter top 5 by token volume this week:

ox alpha's (glm-5.3 flash) volume dropped by more than half – but it's still #1. the hype wave cooled, the promo expires in 5 days, and the leaderboard reshuffled underneath

1. glm 5.3 flash – 11.9t
2. gpt-5.6 luna – 11.6t
3. deepseek v4 flash – 11.3t
4. tencent hy4 preview – 11t
5. tencent hy3 – 5.25t

pricing: $0.05 to $0.834 per million input tokens. hy4 preview debuted at #4 with 11t tokens – tencent's 770b moe at $0.834/m input, the priciest model in the top 5.

biggest movers: luna surged from #7 to #2 with +137%. minimax m3 free appeared at #7 with +907%. mimo-v2.5 dropped from #3 to #8.

the post-hype leaderboard is more interesting than the hype itself – four models neck and neck, and tencent just proved builders will pay 16x the flash price for agent-grade reasoning.

Leaderboard table of top 10 models by token usage on OpenRouter: glm-5.3 flash 11.9T at #1, gpt-5.6 luna 11.6T, deepseek 11.3T, tencent hy4 11T.
Portrait of Alina Fomine, author at thehype.news

Alina Fomine

Author at thehype, covering ai and the builders behind it. Alina works across media, partnerships, and ecosystem growth at /function1 – the ai conference connecting startups, founders, and global tech leaders.

More articles by Alina Fomine →
ON AIR · RADIO.THEHYPE.NEWS ↗ ai news radio — 24/7