fix(pricing): add missing DEFAULT_PRICING entries for current models (#266)

CHANGELOG.md

@@ -9,6 +9,16 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Fixed
 
+- **Missing pricing for current models costed them at $0** — several current
+  models had no `DEFAULT_PRICING` entry, so `get_pricing` returned `None` and
+  they were silently costed at $0 in the Token Usage Summary / cost breakdown
+  (dotted version suffixes like `claude-opus-4.8` are not bridged by the
+  `-`-delimited fuzzy fallback). Added entries for `claude-opus-4.7`,
+  `claude-opus-4.8`, `claude-sonnet-5`, `gpt-5.3-codex`, `gpt-5.4`, `gpt-5.5`,
+  `gpt-5-mini`, `gpt-5.4-mini`, and `gemini-3.5-flash`. (The related
+  sub-workflow usage under-reporting in the same issue was already fixed in
+  [#212](https://github.com/microsoft/conductor/pull/212).)
+  ([#266](https://github.com/microsoft/conductor/issues/266))
 - **For-each dive-in worked only for finished items** — in the web dashboard's
   for-each group detail panel, the per-item "Dive into subworkflow" control was
   nested inside the row's expand/collapse `<button>`, which is `disabled` while

src/conductor/engine/pricing.py

@@ -69,14 +69,27 @@ class ModelPricing:
     "gpt-4.1-mini": ModelPricing(input_per_mtok=0.15, output_per_mtok=0.60),
     "gpt-4": ModelPricing(input_per_mtok=30.00, output_per_mtok=60.00),
     "gpt-3.5-turbo": ModelPricing(input_per_mtok=0.50, output_per_mtok=1.50),
+    # GPT-5.x Series (uniform standard rate; mini variants at the mini tier)
+    "gpt-5.5": ModelPricing(input_per_mtok=2.00, output_per_mtok=8.00),
+    "gpt-5.4": ModelPricing(input_per_mtok=2.00, output_per_mtok=8.00),
+    "gpt-5.3-codex": ModelPricing(input_per_mtok=2.00, output_per_mtok=8.00),
     "gpt-5.2": ModelPricing(input_per_mtok=2.00, output_per_mtok=8.00),
     "gpt-5.1": ModelPricing(input_per_mtok=2.00, output_per_mtok=8.00),
+    "gpt-5.4-mini": ModelPricing(input_per_mtok=0.15, output_per_mtok=0.60),
+    "gpt-5-mini": ModelPricing(input_per_mtok=0.15, output_per_mtok=0.60),
     # O-series
     "o1": ModelPricing(input_per_mtok=15.00, output_per_mtok=60.00),
     "o1-mini": ModelPricing(input_per_mtok=3.00, output_per_mtok=12.00),
     "o1-preview": ModelPricing(input_per_mtok=15.00, output_per_mtok=60.00),
     "o3-mini": ModelPricing(input_per_mtok=1.10, output_per_mtok=4.40),
-    # Claude 4.5 Series (newest)
+    # Claude 5 Series (newest)
+    "claude-sonnet-5": ModelPricing(
+        input_per_mtok=3.00,
+        output_per_mtok=15.00,
+        cache_read_per_mtok=0.30,
+        cache_write_per_mtok=3.75,
+    ),
+    # Claude 4.5 Series
     "claude-opus-4-5": ModelPricing(
         input_per_mtok=5.00,
         output_per_mtok=25.00,
@@ -133,6 +146,19 @@ class ModelPricing:
         cache_read_per_mtok=0.30,
         cache_write_per_mtok=3.75,
     ),
+    # Claude 4.7 / 4.8 Series
+    "claude-opus-4.8": ModelPricing(
+        input_per_mtok=5.00,
+        output_per_mtok=25.00,
+        cache_read_per_mtok=0.50,
+        cache_write_per_mtok=6.25,
+    ),
+    "claude-opus-4.7": ModelPricing(
+        input_per_mtok=5.00,
+        output_per_mtok=25.00,
+        cache_read_per_mtok=0.50,
+        cache_write_per_mtok=6.25,
+    ),
     # Claude 4 Series
     "claude-opus-4": ModelPricing(
         input_per_mtok=15.00,
@@ -212,6 +238,10 @@ class ModelPricing:
         input_per_mtok=1.25,
         output_per_mtok=5.00,
     ),
+    "gemini-3.5-flash": ModelPricing(
+        input_per_mtok=0.30,
+        output_per_mtok=2.50,
+    ),
 }
 
 

tests/test_engine/test_pricing.py

@@ -111,6 +111,37 @@ def test_default_pricing_table_has_expected_models(self) -> None:
         for model in expected_models:
             assert model in DEFAULT_PRICING, f"Expected {model} in DEFAULT_PRICING"
 
+    @pytest.mark.parametrize(
+        ("model", "input_per_mtok", "output_per_mtok"),
+        [
+            # GPT-5.x family (standard rate) and mini variants (#266).
+            ("gpt-5.5", 2.00, 8.00),
+            ("gpt-5.4", 2.00, 8.00),
+            ("gpt-5.3-codex", 2.00, 8.00),
+            ("gpt-5-mini", 0.15, 0.60),
+            ("gpt-5.4-mini", 0.15, 0.60),
+            # Claude families (#266).
+            ("claude-opus-4.8", 5.00, 25.00),
+            ("claude-opus-4.7", 5.00, 25.00),
+            ("claude-sonnet-5", 3.00, 15.00),
+            # Gemini (#266).
+            ("gemini-3.5-flash", 0.30, 2.50),
+        ],
+    )
+    def test_current_models_have_exact_pricing(
+        self, model: str, input_per_mtok: float, output_per_mtok: float
+    ) -> None:
+        """Current models resolve via an exact key, not the $0 None fallback (#266).
+
+        These dotted version suffixes are not bridged by the ``-``-delimited
+        fuzzy fallback, so a missing exact key silently costs $0.
+        """
+        assert model in DEFAULT_PRICING, f"Expected {model} in DEFAULT_PRICING"
+        pricing = get_pricing(model)
+        assert pricing is not None, f"{model} unexpectedly returned None pricing"
+        assert pricing.input_per_mtok == input_per_mtok
+        assert pricing.output_per_mtok == output_per_mtok
+
 
 class TestFuzzyMatchWarnings:
     """Tests for the fuzzy-match warning behavior (#137)."""
@@ -141,15 +172,18 @@ def test_unknown_model_does_not_warn(self, caplog: pytest.LogCaptureFixture) ->
         assert caplog.records == []
 
     def test_cross_family_name_returns_none_no_warn(self, caplog: pytest.LogCaptureFixture) -> None:
-        # Repro from #137: these names share a textual prefix with claude-opus-4
-        # but are different model families. The delimiter check should reject
-        # the match entirely (returning None and degrading gracefully) rather
-        # than silently inheriting claude-opus-4's pricing/context-window.
+        # Repro from #137: a dotted version suffix shares a textual prefix with
+        # the dashed base key claude-opus-4 but is a different family. The
+        # delimiter check must reject the match entirely (returning None and
+        # degrading gracefully) rather than silently inheriting claude-opus-4's
+        # pricing/context-window. A deliberately synthetic version (4.123) is
+        # used so this stays absent from DEFAULT_PRICING as real dotted models
+        # (4.6/4.7/4.8/…) get added over time.
         with caplog.at_level("WARNING", logger="conductor.engine.pricing"):
-            assert get_pricing("claude-opus-4.7") is None
-            assert get_pricing("claude-opus-4.7-high") is None
-            assert get_pricing("claude-opus-4.7-xhigh") is None
-            assert get_pricing("claude-opus-4.7-1m-internal") is None
+            assert get_pricing("claude-opus-4.123") is None
+            assert get_pricing("claude-opus-4.123-high") is None
+            assert get_pricing("claude-opus-4.123-xhigh") is None
+            assert get_pricing("claude-opus-4.123-1m-internal") is None
         assert caplog.records == []
 
     def test_versioned_suffix_match_warns(self, caplog: pytest.LogCaptureFixture) -> None: