diff --git a/headroom/pricing/litellm_model_resolution.py b/headroom/pricing/litellm_model_resolution.py
index c006f64d9..0a9242001 100644
--- a/headroom/pricing/litellm_model_resolution.py
+++ b/headroom/pricing/litellm_model_resolution.py
@@ -2,9 +2,15 @@
 
 from __future__ import annotations
 
+import re
 from collections.abc import Callable
 from dataclasses import dataclass
 
+# Vertex AI appends @YYYYMMDD version tags to model names at runtime
+# (e.g. "claude-haiku-4-5@20251001"). LiteLLM's database stores bare
+# names without version suffixes, so we strip the suffix before lookup.
+_VERTEX_VERSION_SUFFIX_RE = re.compile(r"@\d{8}$")
+
 
 @dataclass(frozen=True, slots=True)
 class LiteLLMModelPrefixRule:
@@ -55,16 +61,40 @@ PRICE_LOOKUP_PROVIDER_PREFIXES: tuple[str, ...] = (
 )
 
 
+def _strip_vertex_version_suffix(model: str) -> str:
+    """Strip Vertex @YYYYMMDD version suffix if present."""
+    return _VERTEX_VERSION_SUFFIX_RE.sub("", model)
+
+
 def resolution_candidates(model: str) -> tuple[str, ...]:
     """Return ordered LiteLLM keys to try for cost-per-token resolution."""
     candidates = [model]
-    candidates.extend(
-        candidate
-        for rule in MODEL_PREFIX_RULES
-        for candidate in (rule.candidate_for(model),)
-        if candidate is not None
-    )
-    alias = MODEL_ALIASES.get(model)
+
+    # If the model has a Vertex @YYYYMMDD version suffix, also try the bare
+    # name. Vertex appends these at runtime; LiteLLM stores bare names only.
+    bare = _strip_vertex_version_suffix(model)
+    is_vertex_versioned = bare != model
+    if is_vertex_versioned:
+        candidates.append(bare)
+
+    # Apply prefix rules to both the original and bare name so that e.g.
+    # "anthropic/claude-haiku-4-5" is tried after "claude-haiku-4-5".
+    for m in dict.fromkeys([model, bare]):
+        candidates.extend(
+            candidate
+            for rule in MODEL_PREFIX_RULES
+            for candidate in (rule.candidate_for(m),)
+            if candidate is not None
+        )
+
+    # Only add vertex_ai/ candidates for models with @YYYYMMDD suffix —
+    # these are known Vertex-routed models. Non-versioned models should not
+    # get vertex_ai/ candidates to avoid matching wrong pricing tier.
+    if is_vertex_versioned:
+        for m in dict.fromkeys([model, bare]):
+            candidates.append(f"vertex_ai/{m}")
+
+    alias = MODEL_ALIASES.get(model) or MODEL_ALIASES.get(bare)
     if alias:
         candidates.append(alias)
     return tuple(dict.fromkeys(candidates))
@@ -86,11 +116,27 @@ def unwrapped_model_forms(model: str) -> tuple[str, ...]:
 
 def pricing_lookup_candidates(model: str) -> tuple[str, ...]:
     """Return ordered LiteLLM model_cost keys to try for pricing lookup."""
+    bare = _strip_vertex_version_suffix(model)
+    is_vertex_versioned = bare != model
+
     candidates = [model]
-    candidates.extend(f"{prefix}{model}" for prefix in PRICE_LOOKUP_PROVIDER_PREFIXES)
+    if is_vertex_versioned:
+        candidates.append(bare)
+
+    # Try all provider prefixes for both the original and bare name.
+    for m in dict.fromkeys([model, bare]):
+        candidates.extend(f"{prefix}{m}" for prefix in PRICE_LOOKUP_PROVIDER_PREFIXES)
+
     # Unwrapped forms come after the prefixed ones so existing precedence is
     # unchanged for names that already resolved.
-    candidates.extend(unwrapped_model_forms(model))
+    for m in dict.fromkeys([model, bare]):
+        candidates.extend(unwrapped_model_forms(m))
+
+    # Only add vertex_ai/ candidates for models with @YYYYMMDD suffix.
+    if is_vertex_versioned:
+        for m in dict.fromkeys([model, bare]):
+            candidates.append(f"vertex_ai/{m}")
+
     candidates.extend(
         alias for candidate in tuple(candidates) if (alias := MODEL_ALIASES.get(candidate))
     )
diff --git a/headroom/proxy/cost.py b/headroom/proxy/cost.py
index cff6be4b2..b42ac18b0 100644
--- a/headroom/proxy/cost.py
+++ b/headroom/proxy/cost.py
@@ -213,7 +213,7 @@ def build_prefix_cache_stats(
                 # Match model to provider
                 _openai_prefixes = ("gpt", "o1", "o3", "o4")
                 is_match = (
-                    (provider == "anthropic" and "claude" in model_name)
+                    (provider in ("anthropic", "vertex:anthropic") and "claude" in model_name)
                     or (provider == "openai" and any(p in model_name for p in _openai_prefixes))
                     or (provider == "gemini" and "gemini" in model_name)
                     or (provider == "bedrock" and "claude" in model_name)
diff --git a/tests/test_pricing_litellm_model_resolution.py b/tests/test_pricing_litellm_model_resolution.py
index c2098a89b..dcfb21485 100644
--- a/tests/test_pricing_litellm_model_resolution.py
+++ b/tests/test_pricing_litellm_model_resolution.py
@@ -3,6 +3,7 @@ from __future__ import annotations
 from headroom.pricing.litellm_model_resolution import (
     MODEL_ALIASES,
     LiteLLMModelPrefixRule,
+    _strip_vertex_version_suffix,
     pricing_lookup_candidates,
     resolution_candidates,
     resolve_litellm_model_name,
@@ -21,11 +22,11 @@ def test_resolution_candidates_try_bare_then_matching_prefix_then_alias() -> Non
     assert resolution_candidates("MiniMax-M3") == ("MiniMax-M3", "minimax/MiniMax-M3")
 
     retired = "claude-3-5-sonnet-20241022"
-    assert resolution_candidates(retired) == (
-        retired,
-        f"anthropic/{retired}",
-        MODEL_ALIASES[retired],
-    )
+    candidates = resolution_candidates(retired)
+    assert candidates[0] == retired
+    assert f"anthropic/{retired}" in candidates
+    assert f"vertex_ai/{retired}" not in candidates  # no @YYYYMMDD suffix = not Vertex
+    assert MODEL_ALIASES[retired] in candidates
 
 
 def test_pricing_lookup_candidates_include_provider_prefixes_and_aliases() -> None:
@@ -59,3 +60,64 @@ def test_resolve_litellm_model_name_returns_first_known_candidate() -> None:
 
 def test_resolve_litellm_model_name_returns_original_when_unknown() -> None:
     assert resolve_litellm_model_name("mystery-model", lambda _: False) == "mystery-model"
+
+
+def test_strip_vertex_version_suffix() -> None:
+    assert _strip_vertex_version_suffix("claude-haiku-4-5@20251001") == "claude-haiku-4-5"
+    assert _strip_vertex_version_suffix("claude-opus-4@20250514") == "claude-opus-4"
+    assert _strip_vertex_version_suffix("claude-sonnet-4-6") == "claude-sonnet-4-6"
+    assert _strip_vertex_version_suffix("claude-sonnet-4-20250514") == "claude-sonnet-4-20250514"
+
+
+def test_resolution_candidates_vertex_versioned_models() -> None:
+    # Vertex appends @YYYYMMDD — bare name and vertex_ai/ must be candidates
+    candidates = resolution_candidates("claude-haiku-4-5@20251001")
+    assert "claude-haiku-4-5" in candidates
+    assert "anthropic/claude-haiku-4-5" in candidates
+    assert "vertex_ai/claude-haiku-4-5" in candidates  # vertex_ai/ only for versioned
+
+    candidates = resolution_candidates("claude-opus-4@20250514")
+    assert "claude-opus-4" in candidates
+    assert "anthropic/claude-opus-4" in candidates
+    assert "vertex_ai/claude-opus-4" in candidates
+
+    # Non-versioned names should NOT get vertex_ai/ candidates
+    candidates = resolution_candidates("claude-sonnet-4-6")
+    assert candidates[0] == "claude-sonnet-4-6"
+    assert "vertex_ai/claude-sonnet-4-6" not in candidates
+    assert "anthropic/claude-sonnet-4-6" in candidates
+
+
+def test_pricing_lookup_candidates_vertex_versioned_models() -> None:
+    candidates = pricing_lookup_candidates("claude-haiku-4-5@20251001")
+    # Bare name and vertex_ai/ prefix must both be candidates
+    assert "claude-haiku-4-5" in candidates
+    assert "vertex_ai/claude-haiku-4-5" in candidates
+    assert "anthropic/claude-haiku-4-5" in candidates
+
+    candidates = pricing_lookup_candidates("claude-opus-4@20250514")
+    assert "claude-opus-4" in candidates
+    assert "vertex_ai/claude-opus-4" in candidates
+
+    # Non-versioned names should NOT get vertex_ai/ pricing candidates
+    candidates = pricing_lookup_candidates("claude-sonnet-4-6")
+    assert "vertex_ai/claude-sonnet-4-6" not in candidates
+    assert "anthropic/claude-sonnet-4-6" in candidates
+
+
+def test_vertex_versioned_model_resolves_to_known_key() -> None:
+    # Simulate LiteLLM knowing the bare model name (not the versioned one)
+    known = {"claude-haiku-4-5", "anthropic/claude-sonnet-4-6"}
+    assert (
+        resolve_litellm_model_name("claude-haiku-4-5@20251001", known.__contains__)
+        == "claude-haiku-4-5"
+    )
+    assert (
+        resolve_litellm_model_name("claude-sonnet-4-6", known.__contains__)
+        == "anthropic/claude-sonnet-4-6"
+    )
+    # Unknown versioned model falls back to original
+    assert (
+        resolve_litellm_model_name("claude-unknown@20251001", lambda _: False)
+        == "claude-unknown@20251001"
+    )