diff --git a/Makefile b/Makefile
index 02ba3e36d..160fbc020 100644
--- a/Makefile
+++ b/Makefile
@@ -1,5 +1,5 @@
 # Define the default target now so that it is always the first target
-BUILD_TARGETS = main quantize quantize-stats perplexity embedding vdot train-text-from-scratch convert-llama2c-to-ggml simple save-load-state server embd-input-test gguf llama-bench baby-llama beam_search
+BUILD_TARGETS = main quantize quantize-stats perplexity embedding vdot train-text-from-scratch convert-llama2c-to-ggml simple save-load-state server embd-input-test gguf llama-bench baby-llama beam_search fill-in-middle
 
 # Binaries only useful for tests
 TEST_TARGETS = tests/test-llama-grammar tests/test-grammar-parser tests/test-double-float tests/test-grad0 tests/test-opt tests/test-quantize-fns tests/test-quantize-perf tests/test-sampling tests/test-tokenizer-0-llama tests/test-tokenizer-0-falcon tests/test-tokenizer-1
@@ -427,6 +427,9 @@ baby-llama: examples/baby-llama/baby-llama.cpp ggml.o llama.o common.o $(OBJS)
 beam_search: examples/beam_search/beam_search.cpp build-info.h ggml.o llama.o common.o $(OBJS)
 	$(CXX) $(CXXFLAGS) $(filter-out %.h,$^) -o $@ $(LDFLAGS)
 
+fill-in-middle: examples/fill-in-middle/FIM.c ggml.o llama.o common.o $(OBJS)
+	$(CXX) $(CXXFLAGS) $(filter-out %.h,$^) -o $@ $(LDFLAGS)
+
 ifneq '' '$(or $(filter clean,$(MAKECMDGOALS)),$(LLAMA_METAL))'
 BUILD_TARGETS += metal
 endif
diff --git a/examples/fill-in-middle/CMakeLists.txt b/examples/fill-in-middle/CMakeLists.txt
new file mode 100644
index 000000000..9150d7c1c
--- /dev/null
+++ b/examples/fill-in-middle/CMakeLists.txt
@@ -0,0 +1,5 @@
+set(TARGET FIM)
+add_executable(${TARGET} FIM.c)
+install(TARGETS ${TARGET} RUNTIME)
+target_link_libraries(${TARGET} PRIVATE llama ${CMAKE_THREAD_LIBS_INIT})
+target_compile_features(${TARGET} PRIVATE cxx_std_11)
diff --git a/examples/fill-in-middle/FIM.c b/examples/fill-in-middle/FIM.c
index 5283b7f47..307a31e63 100644
--- a/examples/fill-in-middle/FIM.c
+++ b/examples/fill-in-middle/FIM.c
@@ -8,15 +8,26 @@ For a quick summary of what's going on here, see issue #2818.
 */
 
 
-static inline struct llama_context* create_codellama_fim_context(const char* model_path) {
+static inline struct llama_context* codellama_create_fim_context(const char* model_path, const char** error_message) {
     struct llama_context_params params = llama_context_default_params();
     struct llama_model* model = llama_load_model_from_file(model_path,  params);
+    if (!model) {
+        *error_message = "Failed to load model.";
+        return NULL;
+    }
+
     struct llama_context* context = llama_new_context_with_model(model, params);
+    if (!context) {
+        *error_message = "Failed to create context.";
+        llama_free_model(model);
+        return NULL;
+    }
+
     return context;
 }
 
 static inline char*
-codellama_fill_in_middle(struct llama_context* ctx, const char* prefix, const char* suffix, size_t n_max_tokens, int n_threads, bool spm, char** error_message) {
+codellama_fill_in_middle(struct llama_context* ctx, const char* prefix, const char* suffix, size_t n_max_tokens, int n_threads, bool spm, const char** error_message) {
 
     int num_tokens;
     llama_token* tokens_end = (llama_token*)malloc(sizeof(llama_token) * n_max_tokens);
@@ -45,48 +56,60 @@ codellama_fill_in_middle(struct llama_context* ctx, const char* prefix, const ch
     }
 
     // Append middle token
-    int num_prompt_tokens = (int)(tokens_end - tokens);
     *tokens_end++ = llama_token_middle(ctx);
 
     // Evaluate the LM on the prompt.
-    if (llama_eval(ctx, tokens, num_prompt_tokens, 0, n_threads)) {
+    if (llama_eval(ctx, tokens, (int)(tokens_end - tokens), 0, n_threads)) {
         *error_message = "Failed to evaluate the prompt.";
         free(tokens);
         return NULL;
     }
 
     // Generate tokens until n_max_tokens or the <EOT> token is generated.
-    int num_generated_tokens = 0;
     llama_token* generated_tokens = NULL;
+    size_t num_generated_tokens = 0;
+    int vocab_size = llama_n_vocab(ctx);
     for (size_t i = 0; i < n_max_tokens; i++) {
-        // Evaluate the LM for a single token
-        llama_token* current_token = generated_tokens + num_generated_tokens;
-        if (llama_eval(ctx, current_token, 1, num_generated_tokens, n_threads)) {
+        // Evaluate the LM for a single token, obtaining the logits and probabilities.
+        if (llama_eval(ctx, &generated_tokens[num_generated_tokens], 1, (int)num_generated_tokens, n_threads)) {
             *error_message = "Failed to evaluate the prompt.";
             free(tokens);
             break;
         }
+        float* logits = llama_get_logits(ctx);
 
-        if (*current_token == llama_token_eot(ctx)) {
+        // From the logits, select the most likely token.
+        float highest_log_likelihood = -1;
+        llama_token likeliest_token = -1;
+        for (llama_token token_id = 0; token_id < vocab_size; token_id++) {
+            if (logits[token_id] > highest_log_likelihood) {
+                highest_log_likelihood = logits[token_id];
+                likeliest_token = token_id;
+            }
+        }
+
+        // Don't add the token if it's <EOT>.
+        if (likeliest_token == llama_token_eot(ctx)) {
             break;
         }
 
-        num_generated_tokens++;
+        // Append the token, so it's there for subsequent evaluations.
+        generated_tokens[num_generated_tokens++] = likeliest_token;
     }
 
     // Allocate memory for the final result
     size_t result_length = 0;
     size_t result_capacity = 4096;
-    char* result = (char*)malloc(sizeof(char) * 4096);
+    char* result = (char*)malloc(sizeof(char) * result_capacity);
     if (!result) {
         *error_message = "Failed to allocate memory for result.";
         free(tokens);
         return NULL;
     }
 
-    // Translate tokens to string
+    // Translate tokens to string, growing the allocation if it's too small.
     for (size_t i = 0; i < num_generated_tokens; i++) {
-        int appended = llama_token_to_str(ctx, generated_tokens[i], result, result_capacity - result_length);
+        int appended = llama_token_to_piece(ctx, generated_tokens[i], result, result_capacity - result_length);
         if (appended < 0) {
             // retry the token with a larger buffer
             i--;
@@ -116,20 +139,27 @@ int main(int argc, char** argv) {
         return 1;
     }
 
-    struct llama_context* ctx = create_codellama_fim_context(argv[1]);
-
+    char* model = argv[1];
+    char* prefix = argv[2];
+    char* suffix = argv[3];
     size_t n_max_tokens = atoi(argv[4]);
     int n_threads = atoi(argv[5]);
     bool spm = false;
-    char* error_message = NULL;
-    char* result = codellama_fill_in_middle(ctx, argv[2], argv[3], n_max_tokens, n_threads, spm, &error_message);
+    const char* error_message = NULL;
 
+    struct llama_context* ctx = codellama_create_fim_context(model, &error_message);
     if (error_message) {
         fprintf(stderr, "Error: %s\n", error_message);
         return 1;
     }
 
-    printf("%s\n", result);
+    char* result = codellama_fill_in_middle(ctx, prefix, suffix, n_max_tokens, n_threads, spm, &error_message);
+    if (error_message) {
+        fprintf(stderr, "Error: %s\n", error_message);
+        return 1;
+    }
+
+    printf("%s%s%s\n", prefix, result, suffix);
 
     llama_free(ctx);
 }
diff --git a/llama.cpp b/llama.cpp
index 35960e7f6..b1dae5304 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -960,6 +960,7 @@ struct llama_vocab {
     id linefeed_id = 13;
 
     // codellama FIM special tokens
+    // TODO: load these from the vocabulary.
     id special_prefix_id = 32007;
     id special_middle_id = 32009;
     id special_suffix_id = 32008;