diff --git a/sirius/README.md b/sirius/README.md
index acb23ea77..c9612edb5 100644
--- a/sirius/README.md
+++ b/sirius/README.md
@@ -10,6 +10,8 @@ The benchmark script has been validated on both Lambda Cloud `GH200` instances a
 
 To run the benchmark on AWS, launch an EC2 instance using the `Deep Learning Base AMI with Single CUDA (Ubuntu 22.04)` (x86 version), which includes CUDA preinstalled.
 
+You will also need to make a GPU capacity reservation upfront, e.g. 1 day for ~100 USD.
+
 ### Lambda Cloud
 
-Running the benchmark on Lambda Cloud requires no additional setup.
\ No newline at end of file
+Running the benchmark on Lambda Cloud requires no additional setup.
diff --git a/sirius/results/lambda-GH200.json b/sirius/results/lambda-GH200.json
index 768c623d0..d176f5265 100644
--- a/sirius/results/lambda-GH200.json
+++ b/sirius/results/lambda-GH200.json
@@ -6,7 +6,7 @@
     "proprietary": "no",
     "hardware": "gpu",
     "tuned": "no",
-    "tags": ["C++","column-oriented","embedded","lukewarm-cold-run"],
+    "tags": ["C++","column-oriented","embedded"],
     "load_time": 26.30,
     "data_size": 26903326720,
     "result": [
diff --git a/sirius/results/p5.4xlarge.json b/sirius/results/p5.4xlarge.json
index 0cf36013a..f3a5e40b2 100644
--- a/sirius/results/p5.4xlarge.json
+++ b/sirius/results/p5.4xlarge.json
@@ -6,7 +6,7 @@
     "proprietary": "no",
     "hardware": "gpu",
     "tuned": "no",
-    "tags": ["C++","column-oriented","embedded","lukewarm-cold-run"],
+    "tags": ["C++","column-oriented","embedded"],
     "load_time": 237.18,
     "data_size": 26921938944,
     "result": [
diff --git a/sirius/template.json b/sirius/template.json
index 512372224..3b60099c6 100644
--- a/sirius/template.json
+++ b/sirius/template.json
@@ -6,7 +6,6 @@
   "tags": [
     "C++",
     "column-oriented",
-    "embedded",
-    "lukewarm-cold-run"
+    "embedded"
   ]
 }