From b7882b2434f2d558acf3fc3906772920089fa3aa Mon Sep 17 00:00:00 2001
From: Abiba <abiba@sysloggh.com>
Date: Mon, 25 May 2026 00:31:40 +0000
Subject: [PATCH] fix: reduce 27B Dense context to 192K to free VRAM

RTX 3090 was at 94.9% VRAM at 262K context. Reduced to 192K (196608),
freeing ~2.4GB. VRAM now at 85% with room for active inference.
---
 router/router.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/router/router.py b/router/router.py
index 6e9cfb8..c1a239e 100644
--- a/router/router.py
+++ b/router/router.py
@@ -27,7 +27,7 @@ GPU_MAX_CONCURRENT = {
 # Context window sizes (tokens) — used for compaction signals
 GPU_CONTEXT = {
     "qwen3.6-35B-A3B": 262144,
-    "qwen3.6-27B-code": 262144,
+    "qwen3.6-27B-code": 196608,
     "qwen3.5-9b-vlm": 262144,
 }