diff --git a/src/api/providers/__tests__/cost.test.ts b/src/api/providers/__tests__/cost.test.ts
new file mode 100644
index 00000000000..8bba2f76ac2
--- /dev/null
+++ b/src/api/providers/__tests__/cost.test.ts
@@ -0,0 +1,148 @@
+import { getCost } from "../cost"
+
+describe("getCost", () => {
+	it("should return the correct cost for Bedrock provider with invokedModelId", () => {
+		// For 1000 tokens with 25% input (250) and 75% output (750)
+		// Claude-3-5-sonnet: (0.003/1000 * 250) + (0.015/1000 * 750) = 0.00075 + 0.01125 = 0.012
+		const cost = getCost("bedrock", "test prompt", "gpt-3.5-turbo", 1000, "claude-3-5-sonnet")
+		expect(cost).toBeCloseTo(0.012, 5)
+	})
+
+	it("should return 0 for Bedrock provider without invokedModelId", () => {
+		// Since GPT models are not supported on Bedrock and we've removed the fallback,
+		// this should return 0
+		const cost = getCost("bedrock", "test prompt", "any-model", 1000)
+		expect(cost).toBe(0)
+	})
+
+	it("should return 0 for unknown provider", () => {
+		const cost = getCost("unknown" as any, "test prompt", "gpt-3.5-turbo", 1000)
+		expect(cost).toBe(0)
+	})
+
+	it("should use provided input and output tokens when available", () => {
+		// For specific input (300) and output (700) tokens
+		// Claude-3-5-sonnet: (0.003/1000 * 300) + (0.015/1000 * 700) = 0.0009 + 0.0105 = 0.0114
+		const cost = getCost("bedrock", "test prompt", "gpt-3.5-turbo", 1000, "claude-3-5-sonnet", 300, 700)
+		expect(cost).toBeCloseTo(0.0114, 5)
+	})
+
+	it("should handle cache write and cache read tokens", () => {
+		// For specific input (300), output (700), cache write (200), and cache read (100) tokens
+		// Claude-3-5-sonnet:
+		// Input: (0.003/1000 * 300) = 0.0009
+		// Output: (0.015/1000 * 700) = 0.0105
+		// Cache Write: (0.00375/1000 * 200) = 0.00075
+		// Cache Read: (0.0003/1000 * 100) = 0.00003
+		// Total: 0.0009 + 0.0105 + 0.00075 + 0.00003 = 0.01218
+		const cost = getCost("bedrock", "test prompt", "gpt-3.5-turbo", 1000, "claude-3-5-sonnet", 300, 700, 200, 100)
+		expect(cost).toBeCloseTo(0.01218, 5)
+	})
+
+	it("should handle models without cache pricing", () => {
+		// For specific input (300), output (700), cache write (200), and cache read (100) tokens
+		// Claude-3-opus:
+		// Input: (0.015/1000 * 300) = 0.0045
+		// Output: (0.075/1000 * 700) = 0.0525
+		// Cache Write: (0/1000 * 200) = 0
+		// Cache Read: (0/1000 * 100) = 0
+		// Total: 0.0045 + 0.0525 + 0 + 0 = 0.057
+		const cost = getCost("bedrock", "test prompt", "gpt-3.5-turbo", 1000, "claude-3-opus", 300, 700, 200, 100)
+		expect(cost).toBeCloseTo(0.057, 5)
+	})
+})
+
+describe("getBedrockCost", () => {
+	it("should return the correct cost for claude-3-5-sonnet", () => {
+		// For 1000 tokens with 25% input (250) and 75% output (750)
+		// Claude-3-5-sonnet: (0.003/1000 * 250) + (0.015/1000 * 750) = 0.00075 + 0.01125 = 0.012
+		const cost = getCost("bedrock", "test prompt", "any-model", 1000, "claude-3-5-sonnet")
+		expect(cost).toBeCloseTo(0.012, 5)
+	})
+
+	// GPT model tests removed as they are not supported on Bedrock
+
+	it("should return 0 for unknown invokedModelId", () => {
+		const cost = getCost("bedrock", "test prompt", "any-model", 1000, "unknown-model")
+		expect(cost).toBe(0)
+	})
+
+	it("should return 0 when invokedModelId is not provided", () => {
+		// Since we've removed the fallback to model-based cost calculation,
+		// this should return 0
+		const cost = getCost("bedrock", "test prompt", "any-model", 1000)
+		expect(cost).toBe(0)
+	})
+
+	it("should handle intelligent prompt router ARN format", () => {
+		// Test with a full ARN from an intelligent prompt router
+		// For 1000 tokens with 25% input (250) and 75% output (750)
+		// Claude-3-5-sonnet: (0.003/1000 * 250) + (0.015/1000 * 750) = 0.00075 + 0.01125 = 0.012
+		const cost = getCost(
+			"bedrock",
+			"test prompt",
+			"custom-arn",
+			1000,
+			"arn:aws:bedrock:us-west-2:699475926481:inference-profile/us.anthropic.claude-3-5-sonnet-20240620-v1:0",
+		)
+		expect(cost).toBeCloseTo(0.012, 5)
+	})
+
+	it("should return the correct cost for Amazon Nova Pro", () => {
+		// For 1000 tokens with 25% input (250) and 75% output (750)
+		// Amazon Nova Pro: (0.0008/1000 * 250) + (0.0032/1000 * 750) = 0.0002 + 0.0024 = 0.0026
+		const cost = getCost("bedrock", "test prompt", "any-model", 1000, "amazon.nova-pro")
+		expect(cost).toBeCloseTo(0.0026, 5)
+	})
+
+	it("should return the correct cost for Amazon Nova Micro", () => {
+		// For 1000 tokens with 25% input (250) and 75% output (750)
+		// Amazon Nova Micro: (0.000035/1000 * 250) + (0.00014/1000 * 750) = 0.00000875 + 0.000105 = 0.00011375
+		const cost = getCost("bedrock", "test prompt", "any-model", 1000, "amazon.nova-micro")
+		expect(cost).toBeCloseTo(0.00011375, 8)
+	})
+
+	it("should return the correct cost for Amazon Titan Text Express", () => {
+		// For 1000 tokens with 25% input (250) and 75% output (750)
+		// Amazon Titan Text Express: (0.0002/1000 * 250) + (0.0006/1000 * 750) = 0.00005 + 0.00045 = 0.0005
+		const cost = getCost("bedrock", "test prompt", "any-model", 1000, "amazon.titan-text-express")
+		expect(cost).toBeCloseTo(0.0005, 5)
+	})
+
+	it("should return the correct cost for Amazon Titan Text Lite", () => {
+		// For 1000 tokens with 25% input (250) and 75% output (750)
+		// Amazon Titan Text Lite: (0.00015/1000 * 250) + (0.0002/1000 * 750) = 0.0000375 + 0.00015 = 0.0001875
+		const cost = getCost("bedrock", "test prompt", "any-model", 1000, "amazon.titan-text-lite")
+		expect(cost).toBeCloseTo(0.0001875, 7)
+	})
+
+	it("should return the correct cost for Amazon Titan Text Embeddings", () => {
+		// For embeddings, with the default 1:3 input/output split (250 input, 750 output)
+		// Amazon Titan Text Embeddings: (0.0001/1000 * 250) = 0.000025
+		// Note: Even though embeddings don't have output tokens, the getCost function
+		// still splits tokens using a 1:3 ratio by default
+		const cost = getCost("bedrock", "test prompt", "any-model", 1000, "amazon.titan-text-embeddings")
+		expect(cost).toBeCloseTo(0.000025, 6)
+	})
+
+	it("should return the correct cost for Llama 3.2 (11B)", () => {
+		// For 1000 tokens with 25% input (250) and 75% output (750)
+		// Llama 3.2 (11B): (0.00016/1000 * 250) + (0.00016/1000 * 750) = 0.00004 + 0.00012 = 0.00016
+		const cost = getCost("bedrock", "test prompt", "any-model", 1000, "llama-3.2-11b")
+		expect(cost).toBeCloseTo(0.00016, 6)
+	})
+
+	it("should return the correct cost for Llama 3.2 (90B)", () => {
+		// For 1000 tokens with 25% input (250) and 75% output (750)
+		// Llama 3.2 (90B): (0.00072/1000 * 250) + (0.00072/1000 * 750) = 0.00018 + 0.00054 = 0.00072
+		const cost = getCost("bedrock", "test prompt", "any-model", 1000, "llama-3.2-90b")
+		expect(cost).toBeCloseTo(0.00072, 6)
+	})
+
+	it("should return the correct cost for Llama 3.3 (70B)", () => {
+		// For 1000 tokens with 25% input (250) and 75% output (750)
+		// Llama 3.3 (70B): (0.00072/1000 * 250) + (0.00072/1000 * 750) = 0.00018 + 0.00054 = 0.00072
+		const cost = getCost("bedrock", "test prompt", "any-model", 1000, "llama-3.3-70b")
+		expect(cost).toBeCloseTo(0.00072, 6)
+	})
+})
diff --git a/src/api/providers/bedrock.ts b/src/api/providers/bedrock.ts
index 76d93649604..7e024179a10 100644
--- a/src/api/providers/bedrock.ts
+++ b/src/api/providers/bedrock.ts
@@ -86,6 +86,11 @@ export interface StreamEvent {
 			latencyMs: number
 		}
 	}
+	trace?: {
+		promptRouter?: {
+			invokedModelId?: string
+		}
+	}
 }
 
 export class AwsBedrockHandler extends BaseProvider implements SingleCompletionHandler {
@@ -252,10 +257,49 @@ export class AwsBedrockHandler extends BaseProvider implements SingleCompletionH
 
 				// Handle metadata events first
 				if (streamEvent.metadata?.usage) {
+					// Check if this is a response from an intelligent prompt router
+					const invokedModelId = streamEvent.trace?.promptRouter?.invokedModelId
+
+					// If invokedModelId is present, extract it from the ARN format
+					let modelIdForCost: string | undefined
+					if (invokedModelId) {
+						// Extract the model name from the ARN
+						// Example ARN: arn:aws:bedrock:us-west-2:699475926481:inference-profile/us.anthropic.claude-3-5-sonnet-20240620-v1:0
+						const modelMatch = invokedModelId.match(/\/([^\/]+)(?::|$)/)
+						if (modelMatch && modelMatch[1]) {
+							const modelName = modelMatch[1]
+
+							// Map the model name to the format expected by the cost calculation function
+							if (modelName.includes("claude-3-5-sonnet")) {
+								modelIdForCost = "claude-3-5-sonnet"
+							} else if (modelName.includes("claude-3-sonnet")) {
+								modelIdForCost = "claude-3-sonnet"
+							} else if (modelName.includes("claude-3-opus")) {
+								modelIdForCost = "claude-3-opus"
+							} else if (modelName.includes("claude-3-haiku")) {
+								modelIdForCost = "claude-3-haiku"
+							} else if (modelName.includes("claude-3-5-haiku")) {
+								modelIdForCost = "claude-3-5-haiku"
+							} else if (modelName.includes("claude-3-7-sonnet")) {
+								modelIdForCost = "claude-3-7-sonnet"
+							}
+
+							logger.debug("Extracted model ID from intelligent prompt router", {
+								ctx: "bedrock",
+								originalArn: invokedModelId,
+								extractedModelId: modelIdForCost,
+							})
+						}
+					}
+
+					const inputTokens = streamEvent.metadata.usage.inputTokens || 0
+					const outputTokens = streamEvent.metadata.usage.outputTokens || 0
+
 					yield {
 						type: "usage",
-						inputTokens: streamEvent.metadata.usage.inputTokens || 0,
-						outputTokens: streamEvent.metadata.usage.outputTokens || 0,
+						inputTokens: inputTokens,
+						outputTokens: outputTokens,
+						invokedModelId: modelIdForCost,
 					}
 					continue
 				}
@@ -491,6 +535,22 @@ Please check:
 					supportsPromptCache: false,
 					supportsImages: true,
 				}
+			} else if (arnLower.includes("llama3.3") || arnLower.includes("llama-3.3")) {
+				// Llama 3.3 models
+				modelInfo = {
+					maxTokens: 8192,
+					contextWindow: 128_000,
+					supportsPromptCache: true,
+					supportsImages: true,
+				}
+			} else if (arnLower.includes("llama3.2") || arnLower.includes("llama-3.2")) {
+				// Llama 3.2 models
+				modelInfo = {
+					maxTokens: 8192,
+					contextWindow: 128_000,
+					supportsPromptCache: true,
+					supportsImages: arnLower.includes("90b") || arnLower.includes("11b"),
+				}
 			} else if (arnLower.includes("llama3") || arnLower.includes("llama-3")) {
 				// Llama 3 models typically have 8192 tokens in Bedrock
 				modelInfo = {
@@ -499,6 +559,46 @@ Please check:
 					supportsPromptCache: false,
 					supportsImages: arnLower.includes("90b") || arnLower.includes("11b"),
 				}
+			} else if (arnLower.includes("titan-text-lite")) {
+				// Amazon Titan Text Lite
+				modelInfo = {
+					maxTokens: 4096,
+					contextWindow: 8_000,
+					supportsPromptCache: false,
+					supportsImages: false,
+				}
+			} else if (arnLower.includes("titan-text-express")) {
+				// Amazon Titan Text Express
+				modelInfo = {
+					maxTokens: 4096,
+					contextWindow: 8_000,
+					supportsPromptCache: false,
+					supportsImages: false,
+				}
+			} else if (arnLower.includes("titan-text-embeddings")) {
+				// Amazon Titan Text Embeddings
+				modelInfo = {
+					maxTokens: 8192,
+					contextWindow: 8_000,
+					supportsPromptCache: false,
+					supportsImages: false,
+				}
+			} else if (arnLower.includes("nova-micro")) {
+				// Amazon Nova Micro
+				modelInfo = {
+					maxTokens: 4096,
+					contextWindow: 128_000,
+					supportsPromptCache: false,
+					supportsImages: false,
+				}
+			} else if (arnLower.includes("nova-lite")) {
+				// Amazon Nova Lite
+				modelInfo = {
+					maxTokens: 4096,
+					contextWindow: 128_000,
+					supportsPromptCache: false,
+					supportsImages: false,
+				}
 			} else if (arnLower.includes("nova-pro")) {
 				// Amazon Nova Pro
 				modelInfo = {
diff --git a/src/api/providers/cost.ts b/src/api/providers/cost.ts
new file mode 100644
index 00000000000..041ad5660a4
--- /dev/null
+++ b/src/api/providers/cost.ts
@@ -0,0 +1,298 @@
+export function getCost(
+	provider:
+		| "anthropic"
+		| "bedrock"
+		| "openai"
+		| "openai-native"
+		| "vertex"
+		| "unbound"
+		| "glama"
+		| "deepseek"
+		| "ollama"
+		| "mistral"
+		| "requesty"
+		| "gemini"
+		| "human-relay"
+		| "lmstudio"
+		| "openrouter",
+	prompt: string,
+	model: string,
+	tokens: number,
+	invokedModelId?: string,
+	inputTokens?: number,
+	outputTokens?: number,
+	cacheWriteTokens?: number,
+	cacheReadTokens?: number,
+): number {
+	// For Bedrock, handle the case where we have actual input and output token counts
+	if (provider === "bedrock") {
+		// If we have specific input and output token counts, use them
+		if (inputTokens !== undefined && outputTokens !== undefined) {
+			return getBedrockCost(
+				model,
+				inputTokens,
+				outputTokens,
+				invokedModelId,
+				cacheWriteTokens || 0,
+				cacheReadTokens || 0,
+			)
+		}
+
+		// Otherwise, split the tokens between input and output using a 1:3 ratio
+		// This is a reasonable approximation for most conversational use cases
+		const estimatedInputTokens = Math.round(tokens * 0.25)
+		const estimatedOutputTokens = tokens - estimatedInputTokens
+		return getBedrockCost(
+			model,
+			estimatedInputTokens,
+			estimatedOutputTokens,
+			invokedModelId,
+			cacheWriteTokens || 0,
+			cacheReadTokens || 0,
+		)
+	}
+
+	switch (provider) {
+		case "anthropic":
+			return getAnthropicCost(model, tokens)
+		// ... other provider cases
+		default:
+			return 0
+	}
+}
+
+function getBedrockCost(
+	model: string,
+	inputTokens: number,
+	outputTokens: number,
+	invokedModelId?: string,
+	cacheWriteTokens: number = 0,
+	cacheReadTokens: number = 0,
+): number {
+	if (invokedModelId) {
+		// Extract model name from ARN if applicable
+		let modelIdentifier = invokedModelId
+
+		// Check if invokedModelId is an ARN from an intelligent prompt router
+		if (invokedModelId.startsWith("arn:aws:bedrock:")) {
+			// Example ARN: arn:aws:bedrock:us-west-2:699475926481:inference-profile/us.anthropic.claude-3-5-sonnet-20240620-v1:0
+			const modelMatch = invokedModelId.match(/\/([^\/]+)(?::|$)/)
+			if (modelMatch && modelMatch[1]) {
+				modelIdentifier = modelMatch[1]
+			}
+		}
+
+		// March 11, 2025 US region model costs, US-West-2 where specific region is required, per https://aws.amazon.com/bedrock/pricing/
+		//
+		// Define model costs with separate input, output, cache write, and cache read prices per 1000 tokens
+		const modelCosts = [
+			// Claude models - prices from AWS Bedrock pricing documentation
+			{
+				ids: ["claude-3-7-sonnet", "claude-3.7-sonnet"],
+				inputPrice: 0.003,
+				outputPrice: 0.015,
+				cacheWritePrice: 0.00375,
+				cacheReadPrice: 0.0003,
+			},
+			{
+				ids: ["claude-3-5-sonnet", "claude-3.5-sonnet"],
+				inputPrice: 0.003,
+				outputPrice: 0.015,
+				cacheWritePrice: 0.00375,
+				cacheReadPrice: 0.0003,
+			},
+			{
+				ids: ["claude-3-5-haiku", "claude-3.5-haiku"],
+				inputPrice: 0.0008,
+				outputPrice: 0.004,
+				cacheWritePrice: 0.001,
+				cacheReadPrice: 0.00008,
+			},
+			{
+				ids: ["claude-3-opus", "claude-3.0-opus"],
+				inputPrice: 0.015,
+				outputPrice: 0.075,
+				cacheWritePrice: 0, // Not available for this model
+				cacheReadPrice: 0, // Not available for this model
+			},
+			{
+				ids: ["claude-3-haiku", "claude-3.0-haiku"],
+				inputPrice: 0.00025,
+				outputPrice: 0.00125,
+				cacheWritePrice: 0, // Not available for this model
+				cacheReadPrice: 0, // Not available for this model
+			},
+			{
+				ids: ["claude-3-sonnet", "claude-3.0-sonnet"],
+				inputPrice: 0.003,
+				outputPrice: 0.015,
+				cacheWritePrice: 0, // Not available for this model
+				cacheReadPrice: 0, // Not available for this model
+			},
+			{
+				ids: ["claude-2-1", "claude-2.1"],
+				inputPrice: 0.008,
+				outputPrice: 0.024,
+				cacheWritePrice: 0, // Not available for this model
+				cacheReadPrice: 0, // Not available for this model
+			},
+			{
+				ids: ["claude-2"],
+				inputPrice: 0.008,
+				outputPrice: 0.024,
+				cacheWritePrice: 0, // Not available for this model
+				cacheReadPrice: 0, // Not available for this model
+			},
+			{
+				ids: ["claude-instant"],
+				inputPrice: 0.0008,
+				outputPrice: 0.0024,
+				cacheWritePrice: 0, // Not available for this model
+				cacheReadPrice: 0, // Not available for this model
+			},
+
+			// Note: GPT models removed as they are not supported on Bedrock
+
+			// Llama models
+			{
+				ids: ["llama-3-8b", "llama-3.0-8b"],
+				inputPrice: 0.0001,
+				outputPrice: 0.0002,
+				cacheWritePrice: 0.00005,
+				cacheReadPrice: 0.00001,
+			},
+			{
+				ids: ["llama-3-70b", "llama-3.0-70b"],
+				inputPrice: 0.0003,
+				outputPrice: 0.0006,
+				cacheWritePrice: 0.00015,
+				cacheReadPrice: 0.00003,
+			},
+			// Llama 3.2 models
+			{
+				ids: ["llama-3.2-1b", "llama-3.2-1b-instruct"],
+				inputPrice: 0.0001,
+				outputPrice: 0.0001,
+				cacheWritePrice: 0.00005,
+				cacheReadPrice: 0.00005,
+			},
+			{
+				ids: ["llama-3.2-3b", "llama-3.2-3b-instruct"],
+				inputPrice: 0.00015,
+				outputPrice: 0.00015,
+				cacheWritePrice: 0.000075,
+				cacheReadPrice: 0.000075,
+			},
+			{
+				ids: ["llama-3.2-11b", "llama-3.2-11b-instruct"],
+				inputPrice: 0.00016,
+				outputPrice: 0.00016,
+				cacheWritePrice: 0.00008,
+				cacheReadPrice: 0.00008,
+			},
+			{
+				ids: ["llama-3.2-90b", "llama-3.2-90b-instruct"],
+				inputPrice: 0.00072,
+				outputPrice: 0.00072,
+				cacheWritePrice: 0.00036,
+				cacheReadPrice: 0.00036,
+			},
+			// Llama 3.3 models
+			{
+				ids: ["llama-3.3-70b", "llama-3.3-70b-instruct"],
+				inputPrice: 0.00072,
+				outputPrice: 0.00072,
+				cacheWritePrice: 0.00036,
+				cacheReadPrice: 0.00036,
+			},
+
+			// Amazon Titan models
+			{
+				ids: ["amazon.titan-text-lite", "titan-text-lite"],
+				inputPrice: 0.00015,
+				outputPrice: 0.0002,
+				cacheWritePrice: 0, // Not available for this model
+				cacheReadPrice: 0, // Not available for this model
+			},
+			{
+				ids: ["amazon.titan-text-express", "titan-text-express"],
+				inputPrice: 0.0002,
+				outputPrice: 0.0006,
+				cacheWritePrice: 0, // Updated pricing, cache may not be available
+				cacheReadPrice: 0, // Updated pricing, cache may not be available
+			},
+			{
+				ids: ["amazon.titan-text-embeddings", "titan-text-embeddings"],
+				inputPrice: 0.0001,
+				outputPrice: 0, // Embeddings don't have output tokens
+				cacheWritePrice: 0, // Not available for this model
+				cacheReadPrice: 0, // Not available for this model
+			},
+			{
+				ids: ["amazon.titan-text-embeddings-v2", "titan-text-embeddings-v2"],
+				inputPrice: 0.00002,
+				outputPrice: 0, // Embeddings don't have output tokens
+				cacheWritePrice: 0, // Not available for this model
+				cacheReadPrice: 0, // Not available for this model
+			},
+			// Amazon Nova models
+			{
+				ids: ["amazon.nova-micro", "nova-micro"],
+				inputPrice: 0.000035,
+				outputPrice: 0.00014,
+				cacheWritePrice: 0, // Not available for this model
+				cacheReadPrice: 0, // Not available for this model
+			},
+			{
+				ids: ["amazon.nova-lite", "nova-lite"],
+				inputPrice: 0.00006,
+				outputPrice: 0.00024,
+				cacheWritePrice: 0, // Not available for this model
+				cacheReadPrice: 0, // Not available for this model
+			},
+			{
+				ids: ["amazon.nova-pro", "nova-pro"],
+				inputPrice: 0.0008,
+				outputPrice: 0.0032,
+				cacheWritePrice: 0, // Not available for this model
+				cacheReadPrice: 0, // Not available for this model
+			},
+			{
+				ids: ["amazon.nova-pro-loi", "nova-pro-loi"],
+				inputPrice: 0.001,
+				outputPrice: 0.004,
+				cacheWritePrice: 0, // Not available for this model
+				cacheReadPrice: 0, // Not available for this model
+			},
+		]
+
+		// Find matching model - check if modelIdentifier matches any of the ids in the ids array
+		const matchedModel = modelCosts.find((m) =>
+			m.ids.some((id) => modelIdentifier === id || modelIdentifier.includes(id)),
+		)
+
+		if (matchedModel) {
+			const inputCost = (matchedModel.inputPrice / 1000) * inputTokens
+			const outputCost = (matchedModel.outputPrice / 1000) * outputTokens
+			const cacheWriteCost = (matchedModel.cacheWritePrice / 1000) * cacheWriteTokens
+			const cacheReadCost = (matchedModel.cacheReadPrice / 1000) * cacheReadTokens
+
+			return inputCost + outputCost + cacheWriteCost + cacheReadCost
+		}
+
+		// If we don't have a specific price for this model, log it and return 0
+		console.warn(`Unknown invokedModelId for cost calculation: ${invokedModelId}`)
+		return 0
+	} else {
+		// No fallback model-based cost calculation for Bedrock
+		// GPT models are not supported on Bedrock
+		return 0
+	}
+}
+
+function getAnthropicCost(model: string, tokens: number): number {
+	// Anthropic cost calculation logic
+	return 0
+}
+
+export const parseApiPrice = (price: any) => (price ? parseFloat(price) * 1_000_000 : undefined)
diff --git a/src/api/transform/stream.ts b/src/api/transform/stream.ts
index 97751edd90d..4fb16b425ff 100644
--- a/src/api/transform/stream.ts
+++ b/src/api/transform/stream.ts
@@ -18,4 +18,5 @@ export interface ApiStreamUsageChunk {
 	cacheWriteTokens?: number
 	cacheReadTokens?: number
 	totalCost?: number // openrouter
+	invokedModelId?: string // For intelligent prompt router responses
 }