@karminski3: Bring it on — we excel at competition. Chinese models have dominated OpenRouter's top 6 usage models for last week and this week. It's a complete reversal from July last year, when only DeepSeek and Qwen broke out. Now the top usage is Tencent Hunyuan3's free version (can be configured into Lobster…)

X AI KOLs Timeline News
openrouter chinese-models ranking deepseek tencent xiaomi flash-model

Summary

Chinese AI models dominated the top 6 usage on OpenRouter for last week and this week, with the free version of Tencent Hunyuan3 ranking first, Xiaomi MiMo-V2.5 and DeepSeek-V4-Flash in the lead, reflecting the significant increase in influence of domestic models in the open-source community.

Bring it on — we excel at competition. Chinese models have dominated OpenRouter's top 6 usage models for last week and this week. It's a complete reversal from July last year, when only DeepSeek and Qwen broke out. Now the top usage is Tencent Hunyuan3's free version (can be configured into Lobster). Then there is Xiaomi MiMo-V2.5. I strongly suspect that the consecutive free offerings over the past few months have worked. MiMo's approach is simple, straightforward, and effective: release new models and make them free, cultivating user habits. DeepSeek-V4-Flash is now extremely well-positioned. Most of its usage comes from Agent users. Although the flash model has low pricing, that doesn't mean it earns less; on the contrary, users are more willing to use it, especially relying on long contexts, resulting in high request volumes. When sorted by long context, DeepSeek-V4-Flash is top. The rest are other flash models. An interesting insight: if you decide to target the Agent scenario, your model should definitely be named xxx-flash, because this name has high recognition. It's like the cerebellum decides the brain. > #deepseek #openrouter #flash模型
Original Article
View Cached Full Text

Cached at: 07/15/26, 01:42 AM

Bring It On: We Excel at Competing

Chinese models have directly dominated the top 6 usage models on OpenRouter for last week and this week. Compared to last July, it’s a complete reversal — back then, only DeepSeek and Qwen had broken out.

Now the top usage is the free version of Tencent Hunyuan 3 (can be configured into Lobster).

Next is Xiaomi MiMo-V2.5. I strongly suspect that the consecutive free offerings over the past few months have paid off. The MiMo approach is simple, crude, and effective — every time a new model is released, it’s offered for free, cultivating user habits.

DeepSeek-V4-Flash now has a very precise positioning — most of its usage comes from agent users. Although the Flash model is priced low, that doesn’t mean it makes less money; on the contrary, users feel more confident using it. It particularly relies on long context, with high request volume. When sorted by long context usage, DeepSeek-V4-Flash is at the top. The rest are other Flash models.

An interesting takeaway: if you decide to target the agent scenario, then the model you release must be called xxx-flash, because this name carries strong recognition. It’s almost like the subconscious deciding for the conscious mind.

#deepseek #openrouter #flash模型

Similar Articles

@yyyole: The landscape is changing! The AI 'national team' is entering the race at full speed! China Mobile has launched MoMa, resembling a Chinese version of OpenRouter. It is reportedly the largest AI model aggregation platform (MaaS) in the country?? Officially, the platform integrates 300+ models, covering all mainstream models on the market, achieving centralized token procurement, reducing costs by over 30%...

X AI KOLs Timeline

China Mobile has launched the MoMa platform, acting as a Chinese counterpart to OpenRouter. It aggregates over 300 mainstream AI models, aiming to reduce costs by more than 30% and resource usage by over 50% through centralized procurement.