diff --git a/llama_cpp/llama_cpp.py b/llama_cpp/llama_cpp.py
index 5d5b04e..454fe1e 100644
--- a/llama_cpp/llama_cpp.py
+++ b/llama_cpp/llama_cpp.py
@@ -40,6 +40,7 @@ class llama_token_data(Structure):
 
 llama_token_data_p = POINTER(llama_token_data)
 
+llama_progress_callback = ctypes.CFUNCTYPE(None, c_double, c_void_p)
 
 class llama_context_params(Structure):
     _fields_ = [
@@ -54,6 +55,11 @@ class llama_context_params(Structure):
         ("vocab_only", c_bool),  # only load the vocabulary, no weights
         ("use_mlock", c_bool),  # force system to keep model in RAM
         ("embedding", c_bool),  # embedding mode only
+
+        # called with a progress value between 0 and 1, pass NULL to disable
+        ("progress_callback", llama_progress_callback),
+        # context pointer passed to the progress callback
+        ("progress_callback_user_data", c_void_p),
     ]
 
 
diff --git a/vendor/llama.cpp b/vendor/llama.cpp
index 31572d9..502a400 160000
--- a/vendor/llama.cpp
+++ b/vendor/llama.cpp
@@ -1 +1 @@
-Subproject commit 31572d966531f7d768eb773322016ab78eb6e835
+Subproject commit 502a400192013d3e95ed87b777e8fa3bec45713c