diff --git a/workflowai/core/domain/run.py b/workflowai/core/domain/run.py
index 1db9aed..fbf15ca 100644
--- a/workflowai/core/domain/run.py
+++ b/workflowai/core/domain/run.py
@@ -110,7 +110,7 @@ def format_output(self) -> str:
 
         # Add run information if available
         if self.cost_usd is not None:
-            output.append(f"Cost: $ {self.cost_usd}")
+            output.append(f"Cost: $ {self.cost_usd:.5f}")
         if self.duration_seconds is not None:
             output.append(f"Latency: {self.duration_seconds:.2f}s")
 
diff --git a/workflowai/core/domain/run_test.py b/workflowai/core/domain/run_test.py
index 34c65ce..7dc5b1e 100644
--- a/workflowai/core/domain/run_test.py
+++ b/workflowai/core/domain/run_test.py
@@ -67,7 +67,29 @@ def test_format_output() -> None:
   "message": "hello"
 }
 ==================================================
-Cost: $ 0.001
+Cost: $ 0.00100
+Latency: 1.23s"""
+
+    assert run.format_output() == expected
+
+
+def test_format_output_very_low_cost() -> None:
+    run = Run[_TestOutput](
+        id="test-id",
+        agent_id="agent-1",
+        schema_id=1,
+        output=_TestOutput(message="hello"),
+        duration_seconds=1.23,
+        cost_usd=4.97625e-05,
+    )
+
+    expected = """\nOutput:
+==================================================
+{
+  "message": "hello"
+}
+==================================================
+Cost: $ 0.00005
 Latency: 1.23s"""
 
     assert run.format_output() == expected