diff --git a/autogpt_platform/backend/backend/data/platform_cost.py b/autogpt_platform/backend/backend/data/platform_cost.py
index e43ae147d4..3baa0a58a7 100644
--- a/autogpt_platform/backend/backend/data/platform_cost.py
+++ b/autogpt_platform/backend/backend/data/platform_cost.py
@@ -275,6 +275,42 @@ async def get_platform_cost_dashboard(
         "trackingAmount": True,
     }
 
+    # Build parameterised WHERE clause for the raw SQL percentile/bucket
+    # queries so they honour all active dashboard filters, not just start date.
+    raw_params: list = [start]
+    raw_where_clauses = [
+        '"trackingType" = \'cost_usd\'',
+        '"createdAt" >= $1',
+    ]
+    param_idx = 2  # $1 is already start
+
+    if end is not None:
+        raw_where_clauses.append(f'"createdAt" <= ${param_idx}')
+        raw_params.append(end)
+        param_idx += 1
+
+    if provider is not None:
+        raw_where_clauses.append(f'"provider" = ${param_idx}')
+        raw_params.append(provider.lower())
+        param_idx += 1
+
+    if user_id is not None:
+        raw_where_clauses.append(f'"userId" = ${param_idx}')
+        raw_params.append(user_id)
+        param_idx += 1
+
+    if model is not None:
+        raw_where_clauses.append(f'"model" = ${param_idx}')
+        raw_params.append(model)
+        param_idx += 1
+
+    if block_name is not None:
+        raw_where_clauses.append(f'LOWER("blockName") = LOWER(${param_idx})')
+        raw_params.append(block_name)
+        param_idx += 1
+
+    raw_where = " AND ".join(raw_where_clauses)
+
     # Run all six aggregation queries in parallel.
     (
         by_provider_groups,
@@ -317,7 +353,7 @@ async def get_platform_cost_dashboard(
             },
             count=True,
         ),
-        # Percentile distribution of cost per request.
+        # Percentile distribution of cost per request (respects all filters).
         query_raw_with_schema(
             "SELECT"
             "  percentile_cont(0.5) WITHIN GROUP"
@@ -329,11 +365,10 @@ async def get_platform_cost_dashboard(
             "  percentile_cont(0.99) WITHIN GROUP"
             '    (ORDER BY "costMicrodollars") as p99'
             ' FROM {schema_prefix}"PlatformCostLog"'
-            " WHERE \"trackingType\" = 'cost_usd'"
-            ' AND "createdAt" >= $1',
-            start,
+            f" WHERE {raw_where}",
+            *raw_params,
         ),
-        # Histogram buckets for cost distribution.
+        # Histogram buckets for cost distribution (respects all filters).
         query_raw_with_schema(
             "SELECT"
             "  CASE"
@@ -351,11 +386,10 @@ async def get_platform_cost_dashboard(
             "  END as bucket,"
             "  COUNT(*) as count"
             ' FROM {schema_prefix}"PlatformCostLog"'
-            " WHERE \"trackingType\" = 'cost_usd'"
-            ' AND "createdAt" >= $1'
+            f" WHERE {raw_where}"
             " GROUP BY bucket"
             ' ORDER BY MIN("costMicrodollars")',
-            start,
+            *raw_params,
         ),
     )
 
diff --git a/autogpt_platform/backend/backend/data/platform_cost_test.py b/autogpt_platform/backend/backend/data/platform_cost_test.py
index 1ddf2b3e8a..7649eff524 100644
--- a/autogpt_platform/backend/backend/data/platform_cost_test.py
+++ b/autogpt_platform/backend/backend/data/platform_cost_test.py
@@ -416,6 +416,7 @@ class TestGetPlatformCostDashboard:
         mock_actions.group_by = AsyncMock(side_effect=[[], [], [], []])
         mock_actions.find_many = AsyncMock(return_value=[])
 
+        raw_mock = AsyncMock(side_effect=[[], []])
         with (
             patch(
                 "backend.data.platform_cost.PrismaLog.prisma",
@@ -427,8 +428,7 @@ class TestGetPlatformCostDashboard:
             ),
             patch(
                 "backend.data.platform_cost.query_raw_with_schema",
-                new_callable=AsyncMock,
-                side_effect=[[], []],
+                raw_mock,
             ),
         ):
             await get_platform_cost_dashboard(
@@ -440,6 +440,12 @@ class TestGetPlatformCostDashboard:
         # The where dict passed to the first call should include createdAt
         first_call_kwargs = mock_actions.group_by.call_args_list[0][1]
         assert "createdAt" in first_call_kwargs.get("where", {})
+        # Raw SQL queries should receive provider and user_id as parameters
+        assert raw_mock.await_count == 2
+        raw_call_args = raw_mock.call_args_list[0][0]  # positional args of 1st call
+        raw_params = raw_call_args[1:]  # first arg is the query template
+        assert "openai" in raw_params
+        assert "u1" in raw_params
 
 
 def _make_prisma_log_row(
diff --git a/autogpt_platform/frontend/src/app/(platform)/admin/platform-costs/components/PlatformCostContent.tsx b/autogpt_platform/frontend/src/app/(platform)/admin/platform-costs/components/PlatformCostContent.tsx
index 8e785320dc..636f4f1dd8 100644
--- a/autogpt_platform/frontend/src/app/(platform)/admin/platform-costs/components/PlatformCostContent.tsx
+++ b/autogpt_platform/frontend/src/app/(platform)/admin/platform-costs/components/PlatformCostContent.tsx
@@ -205,7 +205,7 @@ export function PlatformCostContent({ searchParams }: Props) {
       {loading ? (
         <div className="flex flex-col gap-4">
           <div className="grid grid-cols-2 gap-4 md:grid-cols-4">
-            {[...Array(8)].map((_, i) => (
+            {[...Array(12)].map((_, i) => (
               <Skeleton key={i} className="h-20 rounded-lg" />
             ))}
           </div>
@@ -240,7 +240,7 @@ export function PlatformCostContent({ searchParams }: Props) {
                   value={formatMicrodollars(
                     dashboard.avg_cost_microdollars_per_request ?? 0,
                   )}
-                  subtitle="Known cost divided by total requests"
+                  subtitle="Known cost divided by cost-bearing requests"
                 />
                 <SummaryCard
                   label="Avg Input Tokens"
@@ -268,6 +268,13 @@ export function PlatformCostContent({ searchParams }: Props) {
                   )}
                   subtitle="Median cost per request"
                 />
+                <SummaryCard
+                  label="P75 Cost / Request"
+                  value={formatMicrodollars(
+                    dashboard.cost_p75_microdollars ?? 0,
+                  )}
+                  subtitle="75th percentile cost"
+                />
                 <SummaryCard
                   label="P95 Cost / Request"
                   value={formatMicrodollars(