diff --git a/modules/llama_cpp_server.py b/modules/llama_cpp_server.py
index 05c07748..c3a8d105 100644
--- a/modules/llama_cpp_server.py
+++ b/modules/llama_cpp_server.py
@@ -378,6 +378,7 @@ class LlamaServer:
             cmd += ["--gpu-layers", str(shared.args.gpu_layers), "--fit", "off"]
         else:
             cmd += ["--fit", "on"]
+            cmd += ["--fit-ctx", "8192"]
             if shared.args.fit_target:
                 cmd += ["--fit-target", shared.args.fit_target]