diff --git a/expose.cpp b/expose.cpp
index 67d634d56..8f787f108 100644
--- a/expose.cpp
+++ b/expose.cpp
@@ -220,6 +220,14 @@ extern "C"
         return generation_finished;
     }
 
+    float get_prompt_eval_time() {
+        return prompt_eval_time;
+    }
+
+    float get_prompt_process_time() {
+        return prompt_process_time;
+    }
+
     const char* get_pending_output() {
        return gpttype_get_pending_output().c_str();
     }
diff --git a/expose.h b/expose.h
index b74718eb9..2425eabd6 100644
--- a/expose.h
+++ b/expose.h
@@ -54,3 +54,5 @@ extern std::string lora_filename;
 extern std::string lora_base;
 extern std::vector<std::string> generated_tokens;
 extern bool generation_finished;
+extern float prompt_eval_time;
+extern float prompt_process_time;
diff --git a/gpttype_adapter.cpp b/gpttype_adapter.cpp
index 8d49b67b3..09e637161 100644
--- a/gpttype_adapter.cpp
+++ b/gpttype_adapter.cpp
@@ -33,6 +33,8 @@ std::string executable_path = "";
 std::string lora_filename = "";
 std::string lora_base = "";
 bool generation_finished;
+float prompt_process_time;
+float prompt_eval_time;
 std::vector<std::string> generated_tokens;
 
 //return val: 0=fail, 1=(original ggml, alpaca), 2=(ggmf), 3=(ggjt)
@@ -807,6 +809,8 @@ generation_outputs gpttype_generate(const generation_inputs inputs, generation_o
     bool stream_sse = inputs.stream_sse;
 
     generation_finished = false; // Set current generation status
+    prompt_eval_time = 0;
+    prompt_process_time = 0;
     generated_tokens.clear(); // New Generation, new tokens
 
     if (params.repeat_last_n < 1)
@@ -1327,6 +1331,8 @@ generation_outputs gpttype_generate(const generation_inputs inputs, generation_o
     fflush(stdout);
     output.status = 1;
     generation_finished = true;
+    prompt_eval_time = pt2;
+    prompt_process_time = pt1;
     snprintf(output.text, sizeof(output.text), "%s", concat_output.c_str());
 
     return output;
diff --git a/koboldcpp.py b/koboldcpp.py
index 2041b0d24..26db2b432 100644
--- a/koboldcpp.py
+++ b/koboldcpp.py
@@ -151,6 +151,8 @@ def init_library():
     handle.new_token.argtypes = [ctypes.c_int]
     handle.get_stream_count.restype = ctypes.c_int
     handle.has_finished.restype = ctypes.c_bool
+    handle.get_prompt_eval_time.restype = ctypes.c_float
+    handle.get_prompt_process_time.restype = ctypes.c_float
     handle.abort_generate.restype = ctypes.c_bool
     handle.get_pending_output.restype = ctypes.c_char_p
 
@@ -485,7 +487,8 @@ class ServerRequestHandler(http.server.SimpleHTTPRequestHandler):
             newprompt = fullprompt
 
             gen = asyncio.run(self.handle_request(genparams, newprompt, basic_api_flag, kai_sse_stream_flag))
-
+            gen['prompt_process_time'] = handle.get_prompt_process_time()
+            gen['prompt_eval_time'] = handle.get_prompt_eval_time()
             try:
                 self.send_response(200)
                 self.end_headers()