@vellumai/assistant 0.11.7-dev.202609011858.9cc8acb → 0.11.7-dev.202609012016.14eabff

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vellumai/assistant",
-  "version": "0.11.7-dev.202609011858.9cc8acb",
+  "version": "0.11.7-dev.202609012016.14eabff",
   "license": "MIT",
   "type": "module",
   "exports": {

package/src/__tests__/anthropic-provider.test.ts

@@ -3602,6 +3602,8 @@ describe("AnthropicProvider — deprecated sampling params (temperature / top_p
     "claude-sonnet-5",
     "anthropic/claude-sonnet-5",
     "claude-fable-5",
+    "claude-fable-5-1",
+    "anthropic/claude-fable-5.1",
   ]) {
     test(`strips temperature, top_p, and top_k for ${model}`, async () => {
       const provider = new AnthropicProvider("sk-ant-test", model);

package/src/__tests__/model-intents.test.ts

@@ -20,7 +20,7 @@ describe("model intents", () => {
       "claude-haiku-4-5-20251001",
     );
     expect(resolveModelIntent("anthropic", "quality-optimized")).toBe(
-      "claude-fable-5",
+      "claude-fable-5-1",
     );
     expect(resolveModelIntent("anthropic", "vision-optimized")).toBe(
       "claude-opus-4-6",
@@ -47,7 +47,7 @@ describe("model intents", () => {
       "anthropic/claude-haiku-4.5",
     );
     expect(resolveModelIntent("vercel-ai-gateway", "quality-optimized")).toBe(
-      "anthropic/claude-fable-5",
+      "anthropic/claude-fable-5.1",
     );
     expect(resolveModelIntent("vercel-ai-gateway", "vision-optimized")).toBe(
       "anthropic/claude-opus-4.6",

package/src/__tests__/pricing.test.ts

@@ -450,6 +450,19 @@ describe("resolvePricingForUsage", () => {
     expect(result.estimatedCostUsd).toBeCloseTo(57.4, 10);
   });
 
+  test("bills Fable 5.1 cache reads at the catalog rate", () => {
+    const result = resolvePricingForUsage("anthropic", "claude-fable-5-1", {
+      directInputTokens: 0,
+      outputTokens: 0,
+      cacheCreationInputTokens: 0,
+      cacheReadInputTokens: 1_000_000,
+      anthropicCacheCreation: null,
+    });
+
+    expect(result.pricingStatus).toBe("priced");
+    expect(result.estimatedCostUsd).toBeCloseTo(0.25, 10);
+  });
+
   test("returns unpriced with null cost for unknown provider", () => {
     const usage: PricingUsage = {
       directInputTokens: 10,

package/src/providers/anthropic/client.ts

@@ -1056,7 +1056,7 @@ export class AnthropicProvider implements Provider {
         /claude-opus-4-[78]\b/.test(effectiveModel) ||
         /claude-opus-5\b/.test(effectiveModel) ||
         /claude-sonnet-5\b/.test(effectiveModel) ||
-        effectiveModel.startsWith("claude-fable-");
+        /claude-fable-/.test(effectiveModel);
       const mergedOutputConfig = {
         ...(output_config ?? {}),
         ...(effort && effort !== "none" && supportsEffort

package/src/providers/model-catalog.ts

@@ -193,6 +193,24 @@ const RAW_PROVIDER_CATALOG: ProviderCatalogEntry[] = [
       linkLabel: "Open Anthropic Console",
     },
     models: [
+      {
+        id: "claude-fable-5-1",
+        displayName: "Claude Fable 5.1",
+        contextWindowTokens: 1000000,
+        maxOutputTokens: 128000,
+        longContextPricingThresholdTokens: 200000,
+        supportsThinking: true,
+        adaptiveThinkingOnly: true,
+        supportsCaching: true,
+        supportsVision: true,
+        supportsToolUse: true,
+        pricing: {
+          inputPer1mTokens: 10,
+          outputPer1mTokens: 50,
+          cacheWritePer1mTokens: 12.5,
+          cacheReadPer1mTokens: 0.25,
+        },
+      },
       {
         id: "claude-fable-5",
         displayName: "Claude Fable 5",
@@ -1086,6 +1104,24 @@ const RAW_PROVIDER_CATALOG: ProviderCatalogEntry[] = [
       // OpenRouter proxies anthropic/* through Anthropic's Messages API, so
       // prompt caching and cache TTL metadata pass through unchanged and
       // billing matches Anthropic's direct rates.
+      {
+        id: "anthropic/claude-fable-5.1",
+        displayName: "Claude Fable 5.1",
+        contextWindowTokens: 1000000,
+        maxOutputTokens: 128000,
+        longContextPricingThresholdTokens: 200000,
+        supportsThinking: true,
+        adaptiveThinkingOnly: true,
+        supportsCaching: true,
+        supportsVision: true,
+        supportsToolUse: true,
+        pricing: {
+          inputPer1mTokens: 10,
+          outputPer1mTokens: 50,
+          cacheWritePer1mTokens: 12.5,
+          cacheReadPer1mTokens: 0.25,
+        },
+      },
       {
         id: "anthropic/claude-fable-5",
         displayName: "Claude Fable 5",
@@ -1934,6 +1970,24 @@ const RAW_PROVIDER_CATALOG: ProviderCatalogEntry[] = [
       // The gateway proxies anthropic/* through Anthropic's Messages API, so
       // prompt caching and cache TTL metadata pass through unchanged and
       // billing matches Anthropic's direct rates.
+      {
+        id: "anthropic/claude-fable-5.1",
+        displayName: "Claude Fable 5.1",
+        contextWindowTokens: 1000000,
+        maxOutputTokens: 128000,
+        longContextPricingThresholdTokens: 200000,
+        supportsThinking: true,
+        adaptiveThinkingOnly: true,
+        supportsCaching: true,
+        supportsVision: true,
+        supportsToolUse: true,
+        pricing: {
+          inputPer1mTokens: 10,
+          outputPer1mTokens: 50,
+          cacheWritePer1mTokens: 12.5,
+          cacheReadPer1mTokens: 0.25,
+        },
+      },
       {
         id: "anthropic/claude-fable-5",
         displayName: "Claude Fable 5",

package/src/providers/model-intents.ts

@@ -20,7 +20,7 @@ const PROVIDER_MODEL_INTENTS: Record<string, Record<ModelIntent, string>> = {
     balanced: "claude-sonnet-4-6",
     "cost-optimized": "claude-haiku-4-5-20251001",
     "latency-optimized": "claude-haiku-4-5-20251001",
-    "quality-optimized": "claude-fable-5",
+    "quality-optimized": "claude-fable-5-1",
     "vision-optimized": "claude-opus-4-6",
   },
   openai: {
@@ -55,14 +55,14 @@ const PROVIDER_MODEL_INTENTS: Record<string, Record<ModelIntent, string>> = {
     balanced: "anthropic/claude-sonnet-4.6",
     "cost-optimized": "deepseek/deepseek-v4-flash",
     "latency-optimized": "anthropic/claude-haiku-4.5",
-    "quality-optimized": "anthropic/claude-fable-5",
+    "quality-optimized": "anthropic/claude-fable-5.1",
     "vision-optimized": "anthropic/claude-opus-4.6",
   },
   "vercel-ai-gateway": {
     balanced: "anthropic/claude-sonnet-4.6",
     "cost-optimized": "deepseek/deepseek-v4-flash",
     "latency-optimized": "anthropic/claude-haiku-4.5",
-    "quality-optimized": "anthropic/claude-fable-5",
+    "quality-optimized": "anthropic/claude-fable-5.1",
     "vision-optimized": "anthropic/claude-opus-4.6",
   },
 };

package/src/util/pricing.ts

@@ -380,7 +380,8 @@ function calculateUsageCost(
     directInputCost +
     outputCost +
     calculateTokenCost(
-      effectivePricing.inputPer1M * ANTHROPIC_PROMPT_CACHE_MULTIPLIERS.read,
+      effectivePricing.cacheReadPer1M ??
+        effectivePricing.inputPer1M * ANTHROPIC_PROMPT_CACHE_MULTIPLIERS.read,
       usage.cacheReadInputTokens,
     ) +
     calculateTokenCost(