parameters added and parakeet tdt

README.md

@@ -6,10 +6,26 @@ app_file: app.py
 pinned: false
 license: apache-2.0
 tags:
-  - ASR
+  - automatic-speech-recognition
+  - hf-asr-leaderboard
+  - asr
+  - speech
   - speech-recognition
+  - leaderboard
+  - evaluation
+  - benchmark
   - Dutch
   - leaderboard
+models:
+  - openai/whisper-large-v3
+  - CohereLabs/cohere-transcribe-03-2026
+  - Qwen/Qwen3-ASR-1.7B
+  - nvidia/parakeet-tdt-0.6b-v3
+language:
+  - nl
+metrics:
+  - wer
+  - rtf
 ---
 
 # Dutch ASR Leaderboard

app.py

@@ -48,9 +48,11 @@ def build_dataframe(
 
         results = r.get("results", {})
 
+        params = r.get("params_billions")
         row = {
             "Model": model_name,
             "License": r.get("license", "?"),
+            "Parameters (B)": f"{params}" if params is not None else "—",
         }
 
         wer_values = {}
@@ -89,7 +91,7 @@ def build_dataframe(
 
     df = pd.DataFrame(rows)
 
-    front = [c for c in ["Model", "License", "Average WER"] if c in df.columns]
+    front = [c for c in ["Model", "License", "Parameters (B)", "Average WER"] if c in df.columns]
     rest = [c for c in df.columns if c not in front]
     df = df[front + rest]
 
@@ -110,10 +112,7 @@ def build_dataframe(
 def refresh(sort_by: str, show_proprietary: bool, *dataset_flags: bool):
     records = load_results()
     datasets_visible = dict(zip(DATASET_KEYS, dataset_flags))
-    df = build_dataframe(records, sort_by, show_proprietary, datasets_visible)
-    n = len(records)
-    status = f"{n} model{'s' if n != 1 else ''} in results/"
-    return df, status
+    return build_dataframe(records, sort_by, show_proprietary, datasets_visible)
 
 
 ABOUT_TEXT = """
@@ -144,6 +143,13 @@ Local models are benchmarked on **H100 HBM2e GPUs** for consistent performance m
 with gr.Blocks(title="Dutch ASR Leaderboard", theme=gr.themes.Default()) as demo:
     gr.Markdown("# Dutch ASR Leaderboard")
     gr.Markdown("Community leaderboard for Dutch ASR models. Lower WER is better.")
+    gr.Markdown(
+        "> **Note:** Some models may be benchmaxxed — trained or fine-tuned on data "
+        "that overlaps with these test sets. Treat results as indicative, not definitive."
+    )
+    gr.Markdown(
+        "Want to add your model? [Submit on GitHub →](https://github.com/tvosch/Dutch-ASR-leaderboard)"
+    )
 
     with gr.Tabs():
         with gr.Tab("🏆 Leaderboard"):
@@ -171,10 +177,6 @@ with gr.Blocks(title="Dutch ASR Leaderboard", theme=gr.themes.Default()) as demo
                             for key, label in DATASET_LABELS.items()
                         }
 
-                with gr.Column(scale=1):
-                    status_text = gr.Markdown("")
-                    refresh_btn = gr.Button("🔄 Refresh", variant="primary")
-
             leaderboard_table = gr.Dataframe(
                 label="Results",
                 interactive=False,
@@ -188,12 +190,11 @@ with gr.Blocks(title="Dutch ASR Leaderboard", theme=gr.themes.Default()) as demo
 
             # Initial load and event handlers
             _inputs = [sort_by, show_proprietary, *datasets_visible.values()]
-            demo.load(fn=refresh, inputs=_inputs, outputs=[leaderboard_table, status_text])
-            sort_by.change(fn=refresh, inputs=_inputs, outputs=[leaderboard_table, status_text])
-            show_proprietary.change(fn=refresh, inputs=_inputs, outputs=[leaderboard_table, status_text])
+            demo.load(fn=refresh, inputs=_inputs, outputs=[leaderboard_table])
+            sort_by.change(fn=refresh, inputs=_inputs, outputs=[leaderboard_table])
+            show_proprietary.change(fn=refresh, inputs=_inputs, outputs=[leaderboard_table])
             for cb in datasets_visible.values():
-                cb.change(fn=refresh, inputs=_inputs, outputs=[leaderboard_table, status_text])
-            refresh_btn.click(fn=refresh, inputs=_inputs, outputs=[leaderboard_table, status_text])
+                cb.change(fn=refresh, inputs=_inputs, outputs=[leaderboard_table])
 
         with gr.Tab("ℹ️ About"):
             gr.Markdown(ABOUT_TEXT)

results/CohereLabs__cohere-transcribe-03-2026.json

@@ -7,8 +7,9 @@
   "backend_used": "vllm",
   "normalize": true,
   "run_info": {
-    "vllm_version": "0.19.0",
-    "package_version": "asr-nl==0.1.0"
+    "api_mode": "transcriptions",
+    "vllm_version": "0.19.1",
+    "package_version": "0.1.0"
   },
   "results": {
     "fleurs_nl": {

results/Murmel-2026-05-04.json

@@ -7,7 +7,8 @@
   "backend_used": "api",
   "normalize": true,
   "run_info": {
-    "vllm_version": "0.19.1",
+    "api_mode": "murmel",
+    "murmel_version": "0.1.0",
     "package_version": "0.1.0"
   },
   "results": {

results/Reson8.json

@@ -7,8 +7,8 @@
   "backend_used": "api",
   "normalize": true,
   "run_info": {
-    "vllm_version": "0.19.0",
-    "package_version": "asr-nl==0.1.0"
+    "api_mode": "reson8",
+    "package_version": "0.1.0"
   },
   "results": {
     "fleurs_nl": {