From 367fe9a42b286b7a6fb45f1ac2518a9388cdffe9 Mon Sep 17 00:00:00 2001
From: mushuiyu886 <yang.haoyu@xydigit.com>
Date: Mon, 20 Jul 2026 10:56:42 +0800
Subject: [PATCH] fix(chutes): report cache-read pricing from live catalog
 (#111253)

---
 extensions/chutes/models.test.ts | 8 +++++++-
 extensions/chutes/models.ts      | 3 ++-
 2 files changed, 9 insertions(+), 2 deletions(-)

diff --git a/extensions/chutes/models.test.ts b/extensions/chutes/models.test.ts
index 2aa20a5467dd..c8931af7444c 100644
--- a/extensions/chutes/models.test.ts
+++ b/extensions/chutes/models.test.ts
@@ -187,7 +187,7 @@ describe("chutes-models", () => {
             input_modalities: ["text", "image"],
             context_length: 200000,
             max_output_length: 16384,
-            pricing: { prompt: 0.1, completion: 0.2 },
+            pricing: { prompt: 0.1, completion: 0.2, input_cache_read: 0.05 },
           },
           { id: "new-provider/simple-model" },
         ],
@@ -201,6 +201,12 @@ describe("chutes-models", () => {
         const secondModel = requireChutesModel(models, 1);
         expect(firstModel.id).toBe("zai-org/GLM-5-TEE");
         expect(secondModel.reasoning).toBe(true);
+        expect(secondModel.cost).toEqual({
+          input: 0.1,
+          output: 0.2,
+          cacheRead: 0.05,
+          cacheWrite: 0,
+        });
         if (!secondModel.compat) {
           throw new Error("expected Chutes API model compat");
         }
diff --git a/extensions/chutes/models.ts b/extensions/chutes/models.ts
index 43bc1b414f1d..f100c694866b 100644
--- a/extensions/chutes/models.ts
+++ b/extensions/chutes/models.ts
@@ -58,6 +58,7 @@ interface ChutesModelEntry {
   pricing?: {
     prompt?: number;
     completion?: number;
+    input_cache_read?: number;
   };
   [key: string]: unknown;
 }
@@ -127,7 +128,7 @@ export async function discoverChutesModels(accessToken?: string): Promise<ModelD
         cost: {
           input: entry.pricing?.prompt || 0,
           output: entry.pricing?.completion || 0,
-          cacheRead: 0,
+          cacheRead: entry.pricing?.input_cache_read || 0,
           cacheWrite: 0,
         },
         contextWindow: asPositiveSafeInteger(entry.context_length) ?? CHUTES_DEFAULT_CONTEXT_WINDOW,