From 73f82dc57c56d20954f507de8936e8ab7198f39d Mon Sep 17 00:00:00 2001
From: sirius0xdev <sirius0xdev@proton.me>
Date: Mon, 4 May 2026 15:11:43 -0400
Subject: [PATCH] Update rtx6000-vllm.yaml

---
 infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml b/infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml
index 0a4bfcc..bcb0cb4 100644
--- a/infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml
+++ b/infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml
@@ -39,7 +39,7 @@ spec:
             - --tensor-parallel-size=1
             - --tokenizer-mode=hf 
             - --gpu-memory-utilization=0.95
-            - --max-model-len=131072
+            - --max-model-len=262144
             - --enable-auto-tool-choice 
             - --enable-chunked-prefill
             - --max-num-batched-tokens=4096