docs: sync LMSYS SGLang blog cards (#32982)

Co-authored-by: sglang-bot <sglang-bot@users.noreply.github.com>

docs/index.mdx

@@ -90,7 +90,7 @@ It is designed to deliver low-latency and high-throughput inference across a wid
     }}
   >
     <a
-      href="https://lmsys.org/blog/2026-07-29-mxfp8-nvfp4-rl/"
+      href="https://lmsys.org/blog/2026-08-12-qwen3-8-day0-support/"
       target="_blank"
       rel="noopener noreferrer"
       style={{
@@ -111,8 +111,8 @@ It is designed to deliver low-latency and high-throughput inference across a wid
         }}
       >
         <img
-          src="https://lmsys.org/images/blog/mxfp8-nvfp4-rl/NVFP4-dq.drawio.png"
+          src="https://lmsys.org/images/blog/qwen3-8-day0-support/cover-qwen3-8.png"
-          alt="Towards Blackwell-Native 8-bit and 4-bit RL: End-to-End MXFP8 and NVFP4 RL in Miles"
+          alt="SGLang and Miles Add Day-0 Support for Qwen3.8"
           style={{
             width: "100%",
             height: "100%",
@@ -131,7 +131,7 @@ It is designed to deliver low-latency and high-throughput inference across a wid
             fontSize: "0.98rem",
           }}
         >
-          {"Towards Blackwell-Native 8-bit and 4-bit RL: End-to-End MXFP8 and NVFP4 RL in Miles"}
+          {"SGLang and Miles Add Day-0 Support for Qwen3.8"}
         </p>
         <p
           style={{
@@ -140,12 +140,12 @@ It is designed to deliver low-latency and high-throughput inference across a wid
             opacity: 0.75,
           }}
         >
-          {"July 29, 2026"}
+          {"Aug 12, 2026"}
         </p>
       </div>
     </a>
     <a
-      href="https://lmsys.org/blog/2026-07-27-kimi-k3-day0-support/"
+      href="https://lmsys.org/blog/2026-08-11-unified-radix-cache/"
       target="_blank"
       rel="noopener noreferrer"
       style={{
@@ -166,8 +166,8 @@ It is designed to deliver low-latency and high-throughput inference across a wid
         }}
       >
         <img
-          src="https://lmsys.org/images/blog/kimi-k3-day0-support/cover-kimi-k3.png"
+          src="https://lmsys.org/images/blog/unified-radix-cache/image1.svg"
-          alt="SGLang and Miles Add Day-0 Support for Kimi K3"
+          alt="Unified Radix Cache: One Tree for Hybrid Model Prefix Caching"
           style={{
             width: "100%",
             height: "100%",
@@ -186,7 +186,7 @@ It is designed to deliver low-latency and high-throughput inference across a wid
             fontSize: "0.98rem",
           }}
         >
-          {"SGLang and Miles Add Day-0 Support for Kimi K3"}
+          {"Unified Radix Cache: One Tree for Hybrid Model Prefix Caching"}
         </p>
         <p
           style={{
@@ -195,12 +195,12 @@ It is designed to deliver low-latency and high-throughput inference across a wid
             opacity: 0.75,
           }}
         >
-          {"July 27, 2026"}
+          {"August 11, 2026"}
         </p>
       </div>
     </a>
     <a
-      href="https://lmsys.org/blog/2026-07-18-opd-support-in-miles/"
+      href="https://lmsys.org/blog/2026-08-11-nemotron-3-5-lightning/"
       target="_blank"
       rel="noopener noreferrer"
       style={{
@@ -221,8 +221,8 @@ It is designed to deliver low-latency and high-throughput inference across a wid
         }}
       >
         <img
-          src="https://lmsys.org/images/blog/opd-support-in-miles/figure-1.png"
+          src="https://lmsys.org/images/blog/nemotron-3-5-lightning/pinchbench-accuracy-vs-time.png"
-          alt="OPD Support in Miles"
+          alt="SGLang Adds Day-0 Support for NVIDIA Nemotron 3.5 Lightning"
           style={{
             width: "100%",
             height: "100%",
@@ -241,7 +241,7 @@ It is designed to deliver low-latency and high-throughput inference across a wid
             fontSize: "0.98rem",
           }}
         >
-          {"OPD Support in Miles"}
+          {"SGLang Adds Day-0 Support for NVIDIA Nemotron 3.5 Lightning"}
         </p>
         <p
           style={{
@@ -250,12 +250,12 @@ It is designed to deliver low-latency and high-throughput inference across a wid
             opacity: 0.75,
           }}
         >
-          {"July 18, 2026"}
+          {"August 11, 2026"}
         </p>
       </div>
     </a>
     <a
-      href="https://lmsys.org/blog/2026-07-15-inkling-day0-support/"
+      href="https://lmsys.org/blog/2026-08-10-meta-muse-glimmer/"
       target="_blank"
       rel="noopener noreferrer"
       style={{
@@ -276,8 +276,8 @@ It is designed to deliver low-latency and high-throughput inference across a wid
         }}
       >
         <img
-          src="https://lmsys.org/images/blog/inkling-day0-support/inkling-cover.png"
+          src="https://lmsys.org/images/blog/2026-08-10-meta-muse-glimmer/cover-muse-glimmer.png"
-          alt="SGLang and Miles Add Day-0 Support for Inkling, a Frontier Multimodal Model"
+          alt="SGLang Adds Day-0 Support for Muse Glimmer, a Multimodal Model Built for Local Agentic Workflows"
           style={{
             width: "100%",
             height: "100%",
@@ -296,7 +296,7 @@ It is designed to deliver low-latency and high-throughput inference across a wid
             fontSize: "0.98rem",
           }}
         >
-          {"SGLang and Miles Add Day-0 Support for Inkling, a Frontier Multimodal Model"}
+          {"SGLang Adds Day-0 Support for Muse Glimmer, a Multimodal Model Built for Local Agentic Workflows"}
         </p>
         <p
           style={{
@@ -305,12 +305,12 @@ It is designed to deliver low-latency and high-throughput inference across a wid
             opacity: 0.75,
           }}
         >
-          {"July 15, 2026"}
+          {"August 10, 2026"}
         </p>
       </div>
     </a>
     <a
-      href="https://lmsys.org/blog/2026-07-13-glm52-optimization/"
+      href="https://lmsys.org/blog/2026-08-07-hpc-ops-sglang/"
       target="_blank"
       rel="noopener noreferrer"
       style={{
@@ -331,8 +331,8 @@ It is designed to deliver low-latency and high-throughput inference across a wid
         }}
       >
         <img
-          src="https://lmsys.org/images/blog/glm52-optimization/glm52-nvfp4-performance-pareto.png"
+          src="https://lmsys.org/images/blog/hpc-ops-sglang/hpc-ops-sglang-cover.webp"
-          alt="Serving GLM5.2 NVFP4 Agentic Workload with SGLang: Reaching 500 TPS in 2 Weeks"
+          alt="HPC-Ops \u00d7 SGLang: High-Performance Attention, Router GEMM, and MoE Kernels from Tencent Hunyuan"
           style={{
             width: "100%",
             height: "100%",
@@ -351,7 +351,7 @@ It is designed to deliver low-latency and high-throughput inference across a wid
             fontSize: "0.98rem",
           }}
         >
-          {"Serving GLM5.2 NVFP4 Agentic Workload with SGLang: Reaching 500 TPS in 2 Weeks"}
+          {"HPC-Ops \u00d7 SGLang: High-Performance Attention, Router GEMM, and MoE Kernels from Tencent Hunyuan"}
         </p>
         <p
           style={{
@@ -360,12 +360,12 @@ It is designed to deliver low-latency and high-throughput inference across a wid
             opacity: 0.75,
           }}
         >
-          {"July 14, 2026"}
+          {"August 7, 2026"}
         </p>
       </div>
     </a>
     <a
-      href="https://lmsys.org/blog/2026-07-10-rocm-miles-dsv4/"
+      href="https://lmsys.org/blog/2026-08-05-glmImage-optimization/"
       target="_blank"
       rel="noopener noreferrer"
       style={{
@@ -386,8 +386,8 @@ It is designed to deliver low-latency and high-throughput inference across a wid
         }}
       >
         <img
-          src="https://lmsys.org/images/blog/rocm-miles-dsv4/preview.png"
+          src="https://lmsys.org/images/blog/2026-08-05-glmImage-optimization/05-fanout.png"
-          alt="Bringing DeepSeek-V4 Flash RL Training to AMD Instinct MI355X GPUs with Miles"
+          alt="Full-Stack Performance Optimization of AR+DiT in SGL-Diffusion"
           style={{
             width: "100%",
             height: "100%",
@@ -406,7 +406,7 @@ It is designed to deliver low-latency and high-throughput inference across a wid
             fontSize: "0.98rem",
           }}
         >
-          {"Bringing DeepSeek-V4 Flash RL Training to AMD Instinct MI355X GPUs with Miles"}
+          {"Full-Stack Performance Optimization of AR+DiT in SGL-Diffusion"}
         </p>
         <p
           style={{
@@ -415,7 +415,7 @@ It is designed to deliver low-latency and high-throughput inference across a wid
             opacity: 0.75,
           }}
         >
-          {"July 10, 2026"}
+          {"August 05, 2026"}
         </p>
       </div>
     </a>