Tencent Hunyuan Hy4 Preview Tops OpenRouter Weekly Usage Chart

Tencent's Hunyuan Hy4 Preview has taken the top spot on OpenRouter's weekly usage leaderboard, consuming 187 million tokens last week at a cost of about $16.2 and overtaking GLM 5.3. Developers generally agree the result matches their real-world experience—especially fewer output tokens, higher cache hit rates, and lower actual costs—yet the model has drawn strikingly little discussion in the Chinese AI community.

Confirmed

@mariofilhoml shared the OpenRouter weekly data: Hy4 consumed 187 million tokens last week (costing $16.2), versus 34 million for GLM 5.3 in the same period (data truncated; per the post only)
@mariofilhoml and @brunovolpato both said Hy4 Preview topping the weekly chart matches their own trial experience and isn't surprising
Comparing Hy4 with GLM 5.3 in use, @mariofilhoml found that even at high output settings Hy4 uses fewer output tokens, gets higher OpenRouter cache hit rates, and is notably cheaper in actual cost
@cantor8 observed that Hy4 is now the most-used model on OpenRouter, yet almost nobody talks about it amid heated discussion of Chinese models like GLM, Qwen, DeepSeek, and Kimi—a stark contrast

Unconfirmed

@mariofilhoml admitted his judgment that Hy4 and GLM 5.3 are comparable is just a "gut feeling" lacking systematic benchmarks
No public benchmark data exists yet on Hy4's specific capabilities or context performance, and @cantor8's post gave no concrete scores

Why it matters

Topping usage shows Hy4 already has substantial adoption in real developer workflows, with its low cost and token efficiency likely key drivers—useful reference for cost-conscious API users
The severe mismatch between community buzz and actual usage suggests that when tracking Chinese models, one shouldn't rely on discussion volume alone; usage data may reveal overlooked models

2026-09-05 ~ 2026-09-07 · 5 related posts

Primary sources

Tencent Hy4 burns 187M weekly tokens on OpenRouter, topping GLM 5.3 at 34M — mariofilhoml · 2026-09-06 03:50 (UTC+08:00)
User finds Tencent Hunyuan Hy4 cheaper than GLM 5.3 via fewer output tokens, high cache hit rate — mariofilhoml · 2026-09-06 03:43 (UTC+08:00)

Tencent's Hunyuan Hy4 Preview tops OpenRouter weekly rankings — mariofilhoml · 2026-09-05
Tencent's Hy4 Preview tops OpenRouter weekly ranking, and devs are weighing it vs GLM — brunovolpato · 2026-09-06
[source] User finds Tencent Hunyuan Hy4 cheaper than GLM 5.3 via fewer output tokens, high cache hit rate — mariofilhoml · 2026-09-06
[source] Tencent Hy4 burns 187M weekly tokens on OpenRouter, topping GLM 5.3 at 34M — mariofilhoml · 2026-09-06
Why is nobody talking about Tencent Hy4, the most-used model on OpenRouter? — cantor8 · 2026-09-07