fix(cost): tier Anthropic cache reads by the whole prompt size

packages/__tests__/cost/modelCostFromRegistry.test.ts

@@ -317,11 +317,40 @@ describe("modelCostBreakdownFromRegistry", () => {
         // Higher tier: $6/M input, $22.50/M output (multipliers inherited from base tier)
         expect(breakdown.inputCost).toBe(250000 * 0.000006);
         expect(breakdown.outputCost).toBe(50000 * 0.0000225);
-        expect(breakdown.cachedInputCost).toBe(10000 * 0.000003 * 0.1);
+        // The cache read belongs to a prompt that is over the threshold, so it
+        // bills at the same higher tier as the input - matching how the Gemini
+        // case below prices its cache read.
+        expect(breakdown.cachedInputCost).toBe(10000 * 0.000006 * 0.1);
+        // Cache writes still use the base tier: calculateModelCostBreakdown
+        // prices those from basePricing directly, for every provider.
         expect(breakdown.cacheWrite5mCost).toBe(5000 * 0.000003 * 1.25);
       }
     });
 
+    it("should tier a Claude cache read by the whole prompt size", () => {
+      // Regression: the cache read was priced from tier 0 regardless of prompt
+      // size, because the anthropic threshold function returned 0 for
+      // cachedInputCost. A mostly-cached long prompt was billed at half rate.
+      const modelUsage: ModelUsage = {
+        input: 10000,
+        output: 100,
+        cacheDetails: {
+          cachedInput: 240000, // prompt totals 250K, over the 200K threshold
+        },
+      };
+
+      const breakdown = modelCostBreakdownFromRegistry({
+        modelUsage,
+        providerModelId: "claude-sonnet-4-20250514",
+        provider: "anthropic" as ModelProviderName,
+      });
+
+      expect(breakdown).not.toBeNull();
+      if (breakdown) {
+        expect(breakdown.cachedInputCost).toBe(240000 * 0.000006 * 0.1);
+      }
+    });
+
     it("should use base tier pricing for Gemini 3 Pro Preview under 200K tokens", () => {
       const modelUsage: ModelUsage = {
         input: 150000, // 150K tokens - under threshold

packages/cost/models/calculate-cost.ts

@@ -142,9 +142,13 @@ function getThresholdValueFunction(provider: ModelProviderName): (usage: ModelUs
         switch (field) {
           case "inputCost":
           case "outputCost":
-            return usage.input + 
-              (usage.cacheDetails?.cachedInput ?? 0) + 
-              (usage.cacheDetails?.write5m ?? 0) + 
+          // Anthropic's long-context tier is chosen by the size of the whole
+          // prompt, so a cache read on a >threshold request is billed at the
+          // same higher tier as the input it belongs to.
+          case "cachedInputCost":
+            return usage.input +
+              (usage.cacheDetails?.cachedInput ?? 0) +
+              (usage.cacheDetails?.write5m ?? 0) +
               (usage.cacheDetails?.write1h ?? 0);
           default:
             return 0;