Add Latest releases landing tab: recent models with their benchmark suites and source links

app.py

@@ -119,6 +119,12 @@ input.ctl-search, input.tbl-filter { padding: 7px 10px; border: 1px solid var(--
         border: 1px solid var(--hair); border-radius: 999px; font-size: 11.5px;
         margin: 2px 4px 2px 0; color: var(--muted); white-space: nowrap; }
 .chip i { width: 8px; height: 8px; border-radius: 50%; display: inline-block; flex: none; }
+.rel-hd { display: flex; align-items: center; gap: 8px; flex-wrap: wrap; }
+.rel-name { font-size: 15px; font-weight: 600; color: var(--ink); }
+.rel-meta { color: var(--muted); font-size: 12.5px; margin-left: auto; }
+.rel-chips { margin-top: 8px; }
+.rel-src { color: var(--muted); font-size: 12.5px; margin-top: 6px; }
+.rel-src a { color: var(--accent); }
 .cat-sec { margin: 10px 0 14px; }
 .cat-hd { display: flex; align-items: center; gap: 7px; font-size: 13.5px;
           font-weight: 600; color: var(--ink); margin-bottom: 6px; }
@@ -460,6 +466,65 @@ def render_hero() -> str:
     )
 
 
+# ---------------------------------------------------------------------------
+# Tab 0: Latest releases (landing view)
+# ---------------------------------------------------------------------------
+def rel_date(d: pd.Timestamp) -> str:
+    days = (pd.Timestamp.now().normalize() - d.normalize()).days
+    if days <= 0:
+        return "today"
+    if days == 1:
+        return "yesterday"
+    if days < 7:
+        return f"{days} days ago"
+    if days < 60:
+        return f"{days // 7} week{'s' if days >= 14 else ''} ago"
+    if days < 365:
+        return f"{days // 30} months ago"
+    return f"{days // 365} year{'s' if days >= 730 else ''} ago"
+
+
+MAX_RELEASE_CHIPS = 16
+
+
+def latest_releases_view(count) -> str:
+    recent = MODELS_DF.sort_values("release_date", ascending=False).head(int(count))
+    bucket_rank = {b: i for i, b in enumerate(CATEGORY_ORDER)}
+    cards = []
+    for m in recent.itertuples():
+        usage = (
+            USAGE_DF[USAGE_DF["model_id"] == m.model_id]
+            .drop_duplicates(subset=["benchmark"])
+            .sort_values(["bucket", "benchmark"], key=lambda s: s.map(bucket_rank) if s.name == "bucket" else s)
+        )
+        if usage.empty:
+            body = '<div class="card-sub">No benchmark data recorded for this model.</div>'
+        else:
+            shown = usage.head(MAX_RELEASE_CHIPS)
+            chips = "".join(
+                f'<span class="chip" title="{esc(r.bucket)} · reported category: {esc(r.category)}">'
+                f'<i style="background:var(--cat-{esc(r.bucket)})"></i>{esc(r.benchmark)}</span>'
+                for r in shown.itertuples()
+            )
+            if len(usage) > MAX_RELEASE_CHIPS:
+                chips += f'<span class="chip">+{len(usage) - MAX_RELEASE_CHIPS} more</span>'
+            src = usage.iloc[0]
+            body = (
+                f'<div class="rel-chips">{chips}</div>'
+                f'<div class="rel-src">{len(usage)} benchmarks · Source: '
+                f'<a href="{esc(src["source_url"])}" target="_blank">{esc(src["source_title"])}</a> '
+                f'({esc(src["source_type"])})</div>'
+            )
+        cards.append(
+            f'<div class="card"><div class="rel-hd">'
+            f'<span class="rel-name">{esc(m.model_id)}</span>'
+            f"{openness_chip(bool(m.is_open))}"
+            f'<span class="rel-meta">{esc(m.lab)} · {m.release_date.date()} · {rel_date(m.release_date)}</span>'
+            f"</div>{body}</div>"
+        )
+    return "".join(cards)
+
+
 # ---------------------------------------------------------------------------
 # Tab 1: Benchmark popularity
 # ---------------------------------------------------------------------------
@@ -775,6 +840,12 @@ DEFAULT_BENCH = "GPQA-Diamond" if "GPQA-Diamond" in ALL_BENCHMARKS else ALL_BENC
 with gr.Blocks(title="LLM Benchmark Usage Explorer") as demo:
     gr.HTML(render_hero())
 
+    with gr.Tab("🆕 Latest releases"):
+        gr.HTML('<p class="tab-note">The most recent model releases, the benchmarks they report, and where the numbers come from.</p>')
+        rel_count_pills = PillGroup(["10", "20", "50"], "10", "Show")
+        rel_out = gr.HTML(latest_releases_view("10"))
+        rel_count_pills.change(latest_releases_view, rel_count_pills, rel_out)
+
     with gr.Tab("📊 Popularity & timeline"):
         gr.HTML('<p class="tab-note">Which benchmarks are used the most, and when did they first show up?</p>')
         with gr.Row():