diff --git a/unsloth/models/llama.py b/unsloth/models/llama.py
index ab90d2cbb1..a337472a3e 100644
--- a/unsloth/models/llama.py
+++ b/unsloth/models/llama.py
@@ -1637,7 +1637,7 @@ class FastLlamaModel:
 
         fast_inference    = False, # uses vLLM
         gpu_memory_utilization = 0.5,
-        float8_kv_cache   = True,
+        float8_kv_cache   = False,
         random_state      = 3407,
         max_lora_rank     = 16,
         disable_log_stats = False,
diff --git a/unsloth/models/loader.py b/unsloth/models/loader.py
index 144863b8da..ad312e0040 100644
--- a/unsloth/models/loader.py
+++ b/unsloth/models/loader.py
@@ -76,7 +76,7 @@ class FastLanguageModel(FastLlamaModel):
 
         fast_inference             = False, # uses vLLM
         gpu_memory_utilization     = 0.5,
-        float8_kv_cache            = True,
+        float8_kv_cache            = False,
         random_state               = 3407,
         max_lora_rank              = 16,
         disable_log_stats          = False,