diff --git a/tools/cmd/benchmark/main.go b/tools/cmd/benchmark/main.go
index 78ac2a040..bb0491db8 100644
--- a/tools/cmd/benchmark/main.go
+++ b/tools/cmd/benchmark/main.go
@@ -12,6 +12,7 @@ import (
 
 	"kitty/tools/cli"
 	"kitty/tools/tty"
+	"kitty/tools/tui/graphics"
 	"kitty/tools/tui/loop"
 	"kitty/tools/utils"
 
@@ -139,6 +140,26 @@ func ascii_with_csi() (r result, err error) {
 	return result{"CSI codes with ASCII text", len(out), duration}, nil
 }
 
+func images() (r result, err error) {
+	g := graphics.GraphicsCommand{}
+	g.SetImageId(12345)
+	g.SetQuiet(graphics.GRT_quiet_silent)
+	g.SetAction(graphics.GRT_action_query)
+	g.SetFormat(graphics.GRT_format_rgba)
+	const dim = 1024
+	g.SetDataWidth(dim)
+	g.SetDataHeight(dim)
+	b := strings.Builder{}
+	b.Grow(4*dim*dim + 256)
+	_ = g.WriteWithPayloadTo(&b, make([]byte, 4*dim*dim))
+	data := b.String()
+	duration, err := benchmark_data(data, default_benchmark_options())
+	if err != nil {
+		return result{}, err
+	}
+	return result{"Images", len(data), duration}, nil
+}
+
 var divs = []time.Duration{
 	time.Duration(1), time.Duration(10), time.Duration(100), time.Duration(1000)}
 
@@ -155,15 +176,15 @@ func round(d time.Duration, digits int) time.Duration {
 }
 
 func present_result(r result, col_width int) {
-	rate := float64(r.data_sz) / float64(r.duration)
-	rate *= 1e3
+	rate := float64(r.data_sz) / r.duration.Seconds()
+	rate /= 1024 * 1024
 	f := fmt.Sprintf("%%-%ds", col_width)
-	fmt.Printf("  "+f+" : %-10v @ \x1b[32m%.1f\x1b[m GiB/s\n", r.desc, round(r.duration, 2), rate)
+	fmt.Printf("  "+f+" : %-10v @ \x1b[32m%.1f\x1b[m MB/s\n", r.desc, round(r.duration, 2), rate)
 }
 
 func all_benchamrks() []string {
 	return []string{
-		"ascii", "csi",
+		"ascii", "csi", "images",
 	}
 }
 
@@ -187,6 +208,13 @@ func main(args []string) (err error) {
 		results = append(results, r)
 	}
 
+	if slices.Index(args, "images") >= 0 {
+		if r, err = images(); err != nil {
+			return err
+		}
+		results = append(results, r)
+	}
+
 	fmt.Print(reset)
 	fmt.Println(
 		"These results measure the time it takes the terminal to fully parse all the data sent to it. Some terminals will not render all the data, skipping frames, thereby \"cheating\" in their results. kitty does render all data.")