diff --git a/llama_cpp/llama_cpp.py b/llama_cpp/llama_cpp.py
index ef16272..5de837f 100644
--- a/llama_cpp/llama_cpp.py
+++ b/llama_cpp/llama_cpp.py
@@ -188,6 +188,7 @@ LLAMA_TOKEN_TYPE_BYTE = 6
 #     LLAMA_FTYPE_MOSTLY_IQ2_XXS       = 19, // except 1d tensors
 #     LLAMA_FTYPE_MOSTLY_IQ2_XS        = 20, // except 1d tensors
 #     LLAMA_FTYPE_MOSTLY_Q2_K_S        = 21, // except 1d tensors
+#     LLAMA_FTYPE_MOSTLY_Q3_K_XS       = 22, // except 1d tensors
 
 #     LLAMA_FTYPE_GUESSED = 1024, // not specified in the model file
 # };
@@ -211,6 +212,7 @@ LLAMA_FTYPE_MOSTLY_Q6_K = 18
 LLAMA_FTYPE_MOSTLY_IQ2_XXS = 19
 LLAMA_FTYPE_MOSTLY_IQ2_XS = 20
 LLAMA_FTYPE_MOSTLY_Q2_K_S = 21
+LLAMA_FTYPE_MOSTLY_Q3_K_XS = 22
 LLAMA_FTYPE_GUESSED = 1024
 
 # enum llama_rope_scaling_type {
diff --git a/vendor/llama.cpp b/vendor/llama.cpp
index 504dc37..6f9939d 160000
--- a/vendor/llama.cpp
+++ b/vendor/llama.cpp
@@ -1 +1 @@
-Subproject commit 504dc37be8446fb09b1ede70300250ad41be32a2
+Subproject commit 6f9939d119b2d004c264952eb510bd106455531e