diff --git a/crates/llm_gateway/src/stream_context.rs b/crates/llm_gateway/src/stream_context.rs
index 4cf2e9c5..253d8cea 100644
--- a/crates/llm_gateway/src/stream_context.rs
+++ b/crates/llm_gateway/src/stream_context.rs
@@ -302,9 +302,11 @@ impl HttpContext for StreamContext {
                             - self.ttft_duration.unwrap().as_millis() as u64)
                             / self.response_tokens as u64;
 
+                        debug!("Time per output token: {} milliseconds", tpot);
                         // Record the time per output token
                         self.metrics.time_per_output_token.record(tpot);
 
+                        debug!("Tokens per second: {}", 1000 / tpot);
                         // Record the tokens per second
                         self.metrics.tokens_per_second.record(1000 / tpot);
                     }
diff --git a/crates/llm_gateway/tests/integration.rs b/crates/llm_gateway/tests/integration.rs
index ea65bfa0..7107b4d2 100644
--- a/crates/llm_gateway/tests/integration.rs
+++ b/crates/llm_gateway/tests/integration.rs
@@ -75,6 +75,8 @@ fn setup_filter(module: &mut Tester, config: &str) -> i32 {
         .expect_metric_creation(MetricType::Gauge, "active_http_calls")
         .expect_metric_creation(MetricType::Counter, "ratelimited_rq")
         .expect_metric_creation(MetricType::Histogram, "time_to_first_token")
+        .expect_metric_creation(MetricType::Histogram, "time_per_output_token")
+        .expect_metric_creation(MetricType::Histogram, "tokens_per_second")
         .expect_metric_creation(MetricType::Histogram, "request_latency")
         .expect_metric_creation(MetricType::Histogram, "output_sequence_length")
         .expect_metric_creation(MetricType::Histogram, "input_sequence_length")