update

Signed-off-by: heheda <zhangch99@outlook.com>

Author: heheda12345

_posts/2025-09-11-qwen3-next.md

@@ -45,7 +45,7 @@ In order to manage state for hybrid models like Qwen3-Next, vLLM automatically t
 
 <p align="center">
 <picture>
-<img src="/assets/figures/qwen3-next/hybrid.png" width="30%">
+<img src="/assets/figures/qwen3-next/hybrid.png" width="100%">
 </picture>
 </p>
 
@@ -82,4 +82,4 @@ This effort was made possible thanks to close collaboration with many partners:
 
 vLLM team members who contributed to this effort are: Jie Li, Kaichao You, Chen Zhang, Simon Mo.
 
-👉 Qwen3-Next is now available in **vLLM**. Try it out today and experience **ultra-efficient long-context inference** with the latest hybrid MoE architecture.
+👉 Qwen3-Next is now available in **vLLM**. Try it out today and experience **ultra-efficient long-context inference** with the latest hybrid MoE architecture.