diff --git a/docs/design/metrics.md b/docs/design/metrics.md
index c7bc85756e..3d108890d7 100644
--- a/docs/design/metrics.md
+++ b/docs/design/metrics.md
@@ -33,7 +33,8 @@ In v1, an extensive set of metrics are exposed via a Prometheus-compatible `/met
 - `vllm:request_prompt_tokens` (Histogram) - Histogram of input prompt token counts.
 - `vllm:request_generation_tokens` (Histogram) - Histogram of generation token counts.
 - `vllm:time_to_first_token_seconds` (Histogram) - Time to first token (TTFT).
-- `vllm:inter_token_latency_seconds` (Histogram) - Inter-token latency.
+- `vllm:inter_token_latency_seconds` (Histogram) - Inter-token latency (time between consecutive streamed outputs).
+- `vllm:request_time_per_output_token_seconds` (Histogram) - Per-request Time Per Output Token (TPOT).
 - `vllm:e2e_request_latency_seconds` (Histogram) - End-to-end request latency.
 - `vllm:request_prefill_time_seconds` (Histogram) - Request prefill time.
 - `vllm:request_decode_time_seconds` (Histogram) - Request decode time.
@@ -49,7 +50,8 @@ The subset of metrics exposed in the Grafana dashboard gives us an indication of
 - `vllm:e2e_request_latency_seconds_bucket` - End to end request latency measured in seconds.
 - `vllm:prompt_tokens` - Prompt tokens.
 - `vllm:generation_tokens` - Generation tokens.
-- `vllm:inter_token_latency_seconds` - Inter-token latency (Time Per Output Token, TPOT) in seconds.
+- `vllm:inter_token_latency_seconds` - Inter-token latency in seconds, measured between consecutive streamed outputs. It approximates TPOT when each output contains exactly one token, but differs from request-level TPOT when an output contains multiple tokens or when aggregation weights differ.
+- `vllm:request_time_per_output_token_seconds` - Per-request Time Per Output Token (TPOT) in seconds, computed as `(end-to-end latency - TTFT) / (number of output tokens - 1)`. It is recorded as zero for requests that generate no more than one token.
 - `vllm:time_to_first_token_seconds` - Time to First Token (TTFT) latency in seconds.
 - `vllm:num_requests_running` (also, `_swapped` and `_waiting`) - Number of requests in the RUNNING, WAITING, and SWAPPED states.
 - `vllm:kv_cache_usage_perc` - Percentage of used cache blocks by vLLM.
@@ -63,6 +65,20 @@ The subset of metrics exposed in the Grafana dashboard gives us an indication of
 
 See [the PR which added this Dashboard](https://github.com/vllm-project/vllm/pull/2316) for interesting and useful background on the choices made here.
 
+`vllm:inter_token_latency_seconds` records one sample per streamed output event
+(the wall-clock gap between successive outputs), whereas
+`vllm:request_time_per_output_token_seconds` is recorded once per finished request
+as `(end-to-end latency - TTFT) / (number of output tokens - 1)`. These two
+metrics differ when an output bundles multiple tokens (e.g. speculative
+decoding) or when mean ITL and mean per-request TPOT are aggregated over
+differing weights. See [Understanding the latency metrics](../benchmarking/cli.md#understanding-the-latency-metrics)
+for the benchmark definitions. Requests that generate no more than one token
+are recorded with a TPOT of zero, while `vllm bench serve` excludes them from
+its TPOT statistics. Therefore, the Prometheus histogram and benchmark TPOT
+statistics can differ. Use `vllm:request_time_per_output_token_seconds` for
+request-level TPOT and `vllm:inter_token_latency_seconds` when you specifically
+want the inter-output latency distribution.
+
 ### Prometheus Client Library
 
 Prometheus support was initially added [using the aioprometheus library](https://github.com/vllm-project/vllm/pull/1890), but a switch was made quickly to [prometheus_client](https://github.com/vllm-project/vllm/pull/2730). The rationale is discussed in both linked PRs.
@@ -238,8 +254,7 @@ statistics relating to that iteration:
   iteration.
 - The prefill intervals for any requests that completed prefill in
   this iteration.
-- The inter-token intervals (Time Per Output Token, TPOT), for all
-  requests included in this iteration.
+- The inter-token intervals, for all requests included in this iteration.
 - The Time-To-First-Token (TTFT) for any requests that completed
   prefill in this iteration. However, we calculate this interval
   relative to when the request was first received by the frontend
diff --git a/examples/observability/prometheus_grafana/grafana.json b/examples/observability/prometheus_grafana/grafana.json
index 1c89d45938..6144f2bfd3 100644
--- a/examples/observability/prometheus_grafana/grafana.json
+++ b/examples/observability/prometheus_grafana/grafana.json
@@ -318,7 +318,7 @@
         "type": "prometheus",
         "uid": "${DS_PROMETHEUS}"
       },
-      "description": "Inter token latency in seconds.",
+      "description": "Inter-token latency in seconds, measured between consecutive streamed outputs. It approximates TPOT when each output contains one token, but differs from request-level TPOT under speculative decoding or varying aggregation weights.",
       "fieldConfig": {
         "defaults": {
           "color": {
@@ -479,6 +479,172 @@
       "title": "Inter Token Latency",
       "type": "timeseries"
     },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "${DS_PROMETHEUS}"
+      },
+      "description": "Per-request Time Per Output Token (TPOT) in seconds, computed as (end-to-end latency - TTFT) divided by the number of output tokens minus one. Requests with no more than one output token are recorded as zero; vllm bench serve excludes these requests from its TPOT statistics.",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "custom": {
+            "axisBorderShow": false,
+            "axisCenteredZero": false,
+            "axisColorMode": "text",
+            "axisLabel": "",
+            "axisPlacement": "auto",
+            "barAlignment": 0,
+            "barWidthFactor": 0.6,
+            "drawStyle": "line",
+            "fillOpacity": 0,
+            "gradientMode": "none",
+            "hideFrom": {
+              "legend": false,
+              "tooltip": false,
+              "viz": false
+            },
+            "insertNulls": false,
+            "lineInterpolation": "linear",
+            "lineWidth": 1,
+            "pointSize": 5,
+            "scaleDistribution": {
+              "type": "linear"
+            },
+            "showPoints": "auto",
+            "spanNulls": false,
+            "stacking": {
+              "group": "A",
+              "mode": "none"
+            },
+            "thresholdsStyle": {
+              "mode": "off"
+            }
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              },
+              {
+                "color": "red",
+                "value": 80
+              }
+            ]
+          },
+          "unit": "s"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 48
+      },
+      "id": 17,
+      "options": {
+        "legend": {
+          "calcs": [],
+          "displayMode": "list",
+          "placement": "bottom",
+          "showLegend": true
+        },
+        "tooltip": {
+          "mode": "single",
+          "sort": "none"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "${DS_PROMETHEUS}"
+          },
+          "disableTextWrap": false,
+          "editorMode": "builder",
+          "expr": "histogram_quantile(0.99, sum by(le) (rate(vllm:request_time_per_output_token_seconds_bucket{model_name=\"$model_name\"}[$__rate_interval])))",
+          "fullMetaSearch": false,
+          "includeNullMetadata": false,
+          "instant": false,
+          "legendFormat": "P99",
+          "range": true,
+          "refId": "A",
+          "useBackend": false
+        },
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "${DS_PROMETHEUS}"
+          },
+          "disableTextWrap": false,
+          "editorMode": "builder",
+          "expr": "histogram_quantile(0.95, sum by(le) (rate(vllm:request_time_per_output_token_seconds_bucket{model_name=\"$model_name\"}[$__rate_interval])))",
+          "fullMetaSearch": false,
+          "hide": false,
+          "includeNullMetadata": false,
+          "instant": false,
+          "legendFormat": "P95",
+          "range": true,
+          "refId": "B",
+          "useBackend": false
+        },
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "${DS_PROMETHEUS}"
+          },
+          "disableTextWrap": false,
+          "editorMode": "builder",
+          "expr": "histogram_quantile(0.9, sum by(le) (rate(vllm:request_time_per_output_token_seconds_bucket{model_name=\"$model_name\"}[$__rate_interval])))",
+          "fullMetaSearch": false,
+          "hide": false,
+          "includeNullMetadata": false,
+          "instant": false,
+          "legendFormat": "P90",
+          "range": true,
+          "refId": "C",
+          "useBackend": false
+        },
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "${DS_PROMETHEUS}"
+          },
+          "disableTextWrap": false,
+          "editorMode": "builder",
+          "expr": "histogram_quantile(0.5, sum by(le) (rate(vllm:request_time_per_output_token_seconds_bucket{model_name=\"$model_name\"}[$__rate_interval])))",
+          "fullMetaSearch": false,
+          "hide": false,
+          "includeNullMetadata": false,
+          "instant": false,
+          "legendFormat": "P50",
+          "range": true,
+          "refId": "D",
+          "useBackend": false
+        },
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "${DS_PROMETHEUS}"
+          },
+          "editorMode": "code",
+          "expr": "rate(vllm:request_time_per_output_token_seconds_sum{model_name=\"$model_name\"}[$__rate_interval])\n/\nrate(vllm:request_time_per_output_token_seconds_count{model_name=\"$model_name\"}[$__rate_interval])",
+          "hide": false,
+          "instant": false,
+          "legendFormat": "Mean",
+          "range": true,
+          "refId": "E"
+        }
+      ],
+      "title": "Time Per Output Token (TPOT)",
+      "type": "timeseries"
+    },
     {
       "datasource": {
         "type": "prometheus",