diff --git a/BENCHMARK.md b/BENCHMARK.md
index 3fcb30a..7dad3b4 100644
--- a/BENCHMARK.md
+++ b/BENCHMARK.md
@@ -66,6 +66,10 @@ Ground truth transcripts (`.transcript.json`) with per-word timestamps are hand-
 <img src="benchmark_chart.png" alt="Benchmark comparison on 30s English" width="800">
 </p>
 
+<p align="center">
+<img src="benchmark_scatter.png" alt="Speed vs Accuracy tradeoff" width="700">
+</p>
+
 ### French (16.3 s, 1 speaker, `--language fr`)
 
 | Backend | Policy | Model | RTF | WER | Timestamp MAE |
diff --git a/README.md b/README.md
index d35d569..9853eba 100644
--- a/README.md
+++ b/README.md
@@ -82,7 +82,12 @@ Go to `chrome-extension` for instructions.
 | *[Not recommanded]*  Speaker diarization with Diart | `diart` |
 
 See **Parameters & Configuration** below on how to use them.
-See **[BENCHMARK.md](BENCHMARK.md)** for detailed performance comparisons across all backends.
+
+<p align="center">
+<img src="benchmark_scatter.png" alt="Speed vs Accuracy tradeoff" width="700">
+</p>
+
+See **[BENCHMARK.md](BENCHMARK.md)** for the full benchmark with tables, model size comparison, and more.
 
 
 
diff --git a/benchmark_scatter.png b/benchmark_scatter.png
new file mode 100644
index 0000000..9f62bf3
Binary files /dev/null and b/benchmark_scatter.png differ