diff --git a/README.md b/README.md
index b7315014f..6de4e8374 100644
--- a/README.md
+++ b/README.md
@@ -29,7 +29,7 @@ currently supports:
 * [hugging face hub](https://huggingface.co/models) generative models
 * [replicate](https://replicate.com/) text models
 * [openai api](https://platform.openai.com/docs/introduction) chat & continuation models
-* ggml models like [llama.cpp](https://github.com/ggerganov/llama.cpp)
+* gguf models like [llama.cpp](https://github.com/ggerganov/llama.cpp) version >= 1046
 * .. and many more LLMs!
 
 ## Install:
diff --git a/garak/generators/base.py b/garak/generators/base.py
index 63d78c76a..00045af88 100644
--- a/garak/generators/base.py
+++ b/garak/generators/base.py
@@ -54,7 +54,7 @@ def _call_model(self, prompt: str) -> Union[List[str], str, None]:
         succeed or raise an exception. The @backoff decorator can be helpful
         here - see garak.generators.openai for an example usage.
 
-        Can return None if no reponse was elicited"""
+        Can return None if no response was elicited"""
         raise NotImplementedError
 
     def _pre_generate_hook(self):
diff --git a/garak/generators/ggml.py b/garak/generators/ggml.py
index d13ab467b..c337dfbf9 100644
--- a/garak/generators/ggml.py
+++ b/garak/generators/ggml.py
@@ -1,7 +1,7 @@
 #!/usr/bin/env python3
 """ggml generator support
 
-This generator works with ggml models like llama.cpp.
+This generator works with ggml models in gguf format like llama.cpp.
 
 Put the path to your ggml executable (e.g. "/home/leon/llama.cpp/main") in
 an environment variable named GGML_MAIN_PATH, and pass the path to the
@@ -13,6 +13,7 @@
 """
 
 
+import logging
 import os
 import re
 import subprocess
@@ -20,9 +21,10 @@
 from garak import _config
 from garak.generators.base import Generator
 
+GGUF_MAGIC = bytes([0x47, 0x47, 0x55, 0x46])
 
 class GgmlGenerator(Generator):
-    """Generator interface for ggml models.
+    """Generator interface for ggml models in gguf format.
 
     Set the path to the model as the model name, and put the path to the ggml executable in environment variable GGML_MAIN_PATH.
     """
@@ -34,51 +36,81 @@ class GgmlGenerator(Generator):
     top_p = 0.95
     temperature = 0.8
     exception_on_failure = True
+    first_call = True
 
     generator_family_name = "ggml"
 
+    def command_params(self):
+        return {
+            "-m": self.name,
+            "-n": self.max_tokens,
+            "--repeat-penalty": self.repeat_penalty,
+            "--presence-penalty": self.presence_penalty,
+            "--frequency-penalty": self.frequency_penalty,
+            "--top-k": self.top_k,
+            "--top-p": self.top_p,
+            "--temp": self.temperature,
+            "-s": self.seed,
+        }
+
+
     def __init__(self, name, generations=10):
         self.path_to_ggml_main = os.getenv("GGML_MAIN_PATH")
+        if self.path_to_ggml_main is None:
+            raise RuntimeError("Executable not provided by environment GGML_MAIN_PATH")
+        if not os.path.isfile(self.path_to_ggml_main):
+            raise FileNotFoundError(f"Path provided is not a file: {self.path_to_ggml_main}")
+
+        # this value cannot be `None`, 0 is consistent and `-1` would produce random seeds
+        self.seed = _config.run.seed if _config.run.seed is not None else 0
+
+        # model is a file, validate exists and sanity check file header for supported format
+        if not os.path.isfile(name):
+            raise FileNotFoundError(f"File not found, unable to load model: {name}")
+        else:
+            with open(name, 'rb') as model_file:
+                magic_num = model_file.read(len(GGUF_MAGIC))
+                if magic_num != GGUF_MAGIC:
+                    raise RuntimeError(f"{name} is not in GGUF format")
 
-        self.seed = _config.run.seed
         super().__init__(name, generations=generations)
 
     def _call_model(self, prompt):
         command = [
             self.path_to_ggml_main,
-            "-m",
-            self.name,
-            "-n",
-            self.max_tokens,
-            "--repeat-penalty",
-            self.repeat_penalty,
-            "--presence-penalty",
-            self.presence_penalty,
-            "--frequency-penalty",
-            self.frequency_penalty,
-            "--top-k",
-            self.top_k,
-            "--top-p",
-            self.top_p,
-            "--temp",
-            self.temperature,
-            "-s",
-            self.seed,
             "-p",
             prompt,
         ]
+        # test all params for None type
+        for key, value in self.command_params().items():
+            if value is not None:
+                command.append(key)
+                command.append(value)
         command = [str(param) for param in command]
         if _config.system.verbose > 1:
             print("GGML invoked with", command)
-        result = subprocess.run(
-            command,
-            stdout=subprocess.PIPE,
-            stderr=subprocess.DEVNULL,
-            check=self.exception_on_failure,
-        )
-        output = result.stdout.decode("utf-8")
-        output = re.sub("^" + re.escape(prompt.lstrip()), "", output.lstrip())
-        return output
+        try:
+            result = subprocess.run(
+                command,
+                stdout=subprocess.PIPE,
+                stderr=subprocess.PIPE,
+                check=self.exception_on_failure,
+            )
+            output = result.stdout.decode("utf-8")
+            output = re.sub("^" + re.escape(prompt.lstrip()), "", output.lstrip())
+            self.first_call = False
+            return output
+        except subprocess.CalledProcessError as err:
+            # if this is the first call attempt, raise the exception to indicate
+            # the generator is mis-configured
+            print(err.stderr.decode("utf-8"))
+            logging.error(err.stderr.decode("utf-8"))
+            if self.first_call:
+                raise err
+            return None
+        except Exception as err:
+            logging.error(err)
+            return None
 
 
 default_class = "GgmlGenerator"
diff --git a/pyproject.toml b/pyproject.toml
index fe9e6d836..a9f41b635 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -58,7 +58,8 @@ dependencies = [
   "zalgolib>=0.2.2",
   "ecoji>=0.1.0",
   "deepl==1.17.0",
-  "fschat>=0.2.36"
+  "fschat>=0.2.36",
+  "typing>=3.7,<3.8; python_version<'3.5'"
 ]
 
 [project.urls]
diff --git a/requirements.txt b/requirements.txt
index 0e24f59c5..458a61edb 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -26,4 +26,5 @@ numpy>=1.26.1
 zalgolib>=0.2.2
 ecoji>=0.1.0
 deepl==1.17.0
-fschat>=0.2.36
\ No newline at end of file
+fschat>=0.2.36
+typing>=3.7,<3.8; python_version<'3.5'
\ No newline at end of file