Alina Fomine
The most expensive model in the top 5 just took #1 on OpenRouter
1. Tencent Hy4 preview – 19.1T
2. OpenAI GPT-5.6 Luna – 14.2T
3. DeepSeek V4 Flash – 12.4T
4. z-ai GLM 5.3 Flash – 12.3T
5. Xiaomi MiMo-V2.5 – 5.15T
Pricing: $0.068 to $0.834 per million input tokens. Hy4 preview is 12x more expensive than DeepSeek V4 Flash – and it's routing nearly twice the tokens.
Last week 4 models were within 1T of each other. This week Hy4 broke away by 5T. GLM 5.3 Flash lost #1 as its promo pricing expired. Gemini 3.8 Flash entered at #10 with ↑691% growth.
Builders stopped picking the cheapest model. They started picking the right one per task – and the leaderboard split into lanes.
