diegosouzapw · diegosouzapw · Aug 11, 2026 · Jul 27, 2026
@@ -114,11 +114,11 @@ export const FREE_MODEL_BUDGETS: FreeModelBudget[] = [
   { provider: "bytez", modelId: "Qwen/Qwen2.5-72B-Instruct", displayName: "Qwen/Qwen2.5-72B-Instruct", monthlyTokens: 0, creditTokens: 1000000, freeType: "recurring-credit", poolKey: "bytez", tos: "ambiguous" },
   { provider: "cerebras", modelId: "zai-glm-4.7", displayName: "GLM 4.7", monthlyTokens: 30000000, creditTokens: 0, freeType: "recurring-daily", poolKey: "cerebras", tos: "caution" },
   { provider: "cerebras", modelId: "gpt-oss-120b", displayName: "GPT OSS 120B", monthlyTokens: 30000000, creditTokens: 0, freeType: "recurring-daily", poolKey: "cerebras", tos: "caution" },
-  { provider: "cloudflare-ai", modelId: "@cf/meta/llama-3.3-70b-instruct", displayName: "Llama 3.3 70B (🆓 ~150 resp/day)", monthlyTokens: 30000000, creditTokens: 0, freeType: "recurring-daily", poolKey: "cloudflare-ai", tos: "caution" },
-  { provider: "cloudflare-ai", modelId: "@cf/google/gemma-3-12b-it", displayName: "Gemma 3 12B (🆓)", monthlyTokens: 0, creditTokens: 0, freeType: "recurring-daily", poolKey: "cloudflare-ai", tos: "caution" },
+  // #8717: drop dead Workers AI ids (400/403/410). Keep Neurons/day budget on fp8-fast.
+  { provider: "cloudflare-ai", modelId: "@cf/mistral/mistral-7b-instruct-v0.2-lora", displayName: "Mistral 7B (🆓)", monthlyTokens: 0, creditTokens: 0, freeType: "recurring-daily", poolKey: "cloudflare-ai", tos: "caution" },
   { provider: "cloudflare-ai", modelId: "@cf/qwen/qwen2.5-coder-32b-instruct", displayName: "Qwen 2.5 Coder 32B (🆓)", monthlyTokens: 0, creditTokens: 0, freeType: "recurring-daily", poolKey: "cloudflare-ai", tos: "caution" },
   { provider: "cloudflare-ai", modelId: "@cf/deepseek-ai/deepseek-r1-distill-qwen-32b", displayName: "DeepSeek R1 Distill 32B (🆓)", monthlyTokens: 0, creditTokens: 0, freeType: "recurring-daily", poolKey: "cloudflare-ai", tos: "caution" },
-  { provider: "cloudflare-ai", modelId: "@cf/meta/llama-3.3-70b-instruct-fp8-fast", displayName: "Llama 3.3 70B (FP8 Fast 🆓)", monthlyTokens: 0, creditTokens: 0, freeType: "recurring-daily", poolKey: "cloudflare-ai", tos: "caution" },
+  { provider: "cloudflare-ai", modelId: "@cf/meta/llama-3.3-70b-instruct-fp8-fast", displayName: "Llama 3.3 70B (FP8 Fast 🆓 ~150 resp/day)", monthlyTokens: 30000000, creditTokens: 0, freeType: "recurring-daily", poolKey: "cloudflare-ai", tos: "caution" },
   { provider: "cloudflare-ai", modelId: "@cf/meta/llama-3.2-3b-instruct", displayName: "Llama 3.2 3B (🆓)", monthlyTokens: 0, creditTokens: 0, freeType: "recurring-daily", poolKey: "cloudflare-ai", tos: "caution" },
   { provider: "cloudflare-ai", modelId: "@cf/qwen/qwq-32b", displayName: "QwQ 32B (🆓)", monthlyTokens: 0, creditTokens: 0, freeType: "recurring-daily", poolKey: "cloudflare-ai", tos: "caution" },
   { provider: "cloudflare-ai", modelId: "@cf/zai-org/glm-4.7-flash", displayName: "GLM 4.7 Flash (🆓)", monthlyTokens: 0, creditTokens: 0, freeType: "recurring-daily", poolKey: "cloudflare-ai", tos: "caution" },

@@ -9,14 +9,14 @@ export const cloudflare_aiProvider: RegistryEntry = {
   baseUrl: "https://api.cloudflare.com/client/v4/accounts",
   authType: "apikey",
   authHeader: "bearer",
-  // 10K Neurons/day free: ~150 LLM responses or 500s Whisper audio — global edge
+  // 10K Neurons/day free: ~150 LLM responses or 500s Whisper audio — global edge.
+  // #8717: omit dead ids (llama-3.3-70b-instruct, llama-3.1-8b-instruct,
+  // gemma-3-12b-it, qwen2.5-coder-15b-instruct) — Workers AI returns 400/403/410.
   models: [
-    { id: "@cf/meta/llama-3.3-70b-instruct", name: "Llama 3.3 70B (🆓 ~150 resp/day)" },
-    { id: "@cf/google/gemma-3-12b-it", name: "Gemma 3 12B (🆓)" },
+    { id: "@cf/mistral/mistral-7b-instruct-v0.2-lora", name: "Mistral 7B (🆓)" },
     { id: "@cf/qwen/qwen2.5-coder-32b-instruct", name: "Qwen 2.5 Coder 32B (🆓)" },
     { id: "@cf/deepseek-ai/deepseek-r1-distill-qwen-32b", name: "DeepSeek R1 Distill 32B (🆓)" },
-    // Sweep 2026-06-19: + current Workers AI catalog ids (developers.cloudflare.com/workers-ai/models).
-    { id: "@cf/meta/llama-3.3-70b-instruct-fp8-fast", name: "Llama 3.3 70B (FP8 Fast 🆓)" },
+    { id: "@cf/meta/llama-3.3-70b-instruct-fp8-fast", name: "Llama 3.3 70B (FP8 Fast 🆓 ~150 resp/day)" },
     { id: "@cf/meta/llama-3.2-3b-instruct", name: "Llama 3.2 3B (🆓)" },
     { id: "@cf/qwen/qwq-32b", name: "QwQ 32B (🆓)" },
     {

@@ -0,0 +1,101 @@
+/**
+ * #8717 — Cloudflare Workers AI free catalog must not advertise dead model IDs.
+ *
+ * Four of the original six free-catalog entries return 400/403/410 from
+ * Workers AI (`No such model`, deprecated, or forbidden). #8763 added live
+ * replacements but left the dead IDs in place — free-model routing / combo
+ * builders still pick them and burn quota on guaranteed failures.
+ *
+ * Regression: dead IDs absent from FREE_MODEL_BUDGETS + cloudflare-ai registry;
+ * live replacements present; 30M monthlyTokens budget stays on the live
+ * Llama 3.3 70B FP8 Fast entry (poolKey cloudflare-ai dedupes by max).
+ */
+import test from "node:test";
+import assert from "node:assert/strict";
+import { FREE_MODEL_BUDGETS } from "../../open-sse/config/freeModelCatalog.ts";
+import { cloudflare_aiProvider } from "../../open-sse/config/providers/registry/cloudflare-ai/index.ts";
+
+const DEAD_CLOUDFLARE_MODEL_IDS = [
+  "@cf/meta/llama-3.3-70b-instruct",
+  "@cf/meta/llama-3.1-8b-instruct",
+  "@cf/google/gemma-3-12b-it",
+  "@cf/qwen/qwen2.5-coder-15b-instruct",
+] as const;
+
+const LIVE_CLOUDFLARE_MODEL_IDS = [
+  "@cf/mistral/mistral-7b-instruct-v0.2-lora",
+  "@cf/deepseek-ai/deepseek-r1-distill-qwen-32b",
+  "@cf/qwen/qwen2.5-coder-32b-instruct",
+  "@cf/meta/llama-3.3-70b-instruct-fp8-fast",
+  "@cf/meta/llama-3.2-3b-instruct",
+  "@cf/qwen/qwq-32b",
+  "@cf/zai-org/glm-4.7-flash",
+  "@cf/moonshotai/kimi-k2.6",
+  "@cf/google/gemma-4-26b-a4b-it",
+] as const;
+
+function cloudflareCatalogIds(): string[] {
+  return FREE_MODEL_BUDGETS.filter((m) => m.provider === "cloudflare-ai").map((m) => m.modelId);
+}
+
+function cloudflareRegistryIds(): string[] {
+  assert.equal(cloudflare_aiProvider.id, "cloudflare-ai");
+  return (cloudflare_aiProvider.models ?? []).map((m) => m.id);
+}
+
+test("#8717: FREE_MODEL_BUDGETS must not list dead Cloudflare Workers AI model IDs", () => {
+  const ids = cloudflareCatalogIds();
+  for (const dead of DEAD_CLOUDFLARE_MODEL_IDS) {
+    assert.equal(
+      ids.includes(dead),
+      false,
+      `dead model ${dead} must be removed from freeModelCatalog (still present)`
+    );
+  }
+});
+
+test("#8717: FREE_MODEL_BUDGETS must list live Cloudflare Workers AI replacements", () => {
+  const ids = cloudflareCatalogIds();
+  for (const live of LIVE_CLOUDFLARE_MODEL_IDS) {
+    assert.ok(ids.includes(live), `live model ${live} missing from freeModelCatalog`);
+  }
+});
+
+test("#8717: cloudflare-ai registry must not list dead model IDs", () => {
+  const ids = cloudflareRegistryIds();
+  for (const dead of DEAD_CLOUDFLARE_MODEL_IDS) {
+    assert.equal(
+      ids.includes(dead),
+      false,
+      `dead model ${dead} must be removed from cloudflare-ai registry (still present)`
+    );
+  }
+});
+
+test("#8717: cloudflare-ai registry must list live replacements", () => {
+  const ids = cloudflareRegistryIds();
+  for (const live of LIVE_CLOUDFLARE_MODEL_IDS) {
+    assert.ok(ids.includes(live), `live model ${live} missing from cloudflare-ai registry`);
+  }
+});
+
+test("#8717: 30M monthlyTokens budget moves to live Llama 3.3 70B FP8 Fast", () => {
+  const fp8 = FREE_MODEL_BUDGETS.find(
+    (m) =>
+      m.provider === "cloudflare-ai" && m.modelId === "@cf/meta/llama-3.3-70b-instruct-fp8-fast"
+  );
+  assert.ok(fp8, "fp8-fast entry must exist");
+  assert.equal(
+    fp8.monthlyTokens,
+    30_000_000,
+    "Neurons/day free budget (~30M tok/mo) must stay on the live Llama 3.3 entry"
+  );
+
+  // No other cloudflare-ai catalog row should still claim the 30M (poolKey dedupes by max,
+  // but advertising it on a dead id was the original bug surface).
+  const withBudget = FREE_MODEL_BUDGETS.filter(
+    (m) => m.provider === "cloudflare-ai" && m.monthlyTokens === 30_000_000
+  );
+  assert.equal(withBudget.length, 1);
+  assert.equal(withBudget[0].modelId, "@cf/meta/llama-3.3-70b-instruct-fp8-fast");
+});
@@ -5,7 +5,7 @@ import { CloudflareAIExecutor } from "../../open-sse/executors/cloudflare-ai.ts"
 
 test("CloudflareAIExecutor.buildUrl prefers providerSpecificData.accountId", () => {
   const executor = new CloudflareAIExecutor();
-  const url = executor.buildUrl("@cf/meta/llama-3.3-70b-instruct", true, 0, {
+  const url = executor.buildUrl("@cf/meta/llama-3.3-70b-instruct-fp8-fast", true, 0, {
     accountId: "top-level-id",
     providerSpecificData: { accountId: "provider-id" },
   });
@@ -22,10 +22,10 @@ test("CloudflareAIExecutor.buildUrl falls back to top-level credentials and envi
   process.env.CLOUDFLARE_ACCOUNT_ID = "env-account-id";
 
   try {
-    const fromTopLevel = executor.buildUrl("@cf/meta/llama-3.3-70b-instruct", false, 0, {
+    const fromTopLevel = executor.buildUrl("@cf/meta/llama-3.3-70b-instruct-fp8-fast", false, 0, {
       accountId: "top-level-id",
     });
-    const fromEnv = executor.buildUrl("@cf/meta/llama-3.3-70b-instruct", false, 0, {});
+    const fromEnv = executor.buildUrl("@cf/meta/llama-3.3-70b-instruct-fp8-fast", false, 0, {});
 
     assert.equal(
       fromTopLevel,
@@ -48,7 +48,7 @@ test("CloudflareAIExecutor.buildUrl throws when account ID is missing", () => {
 
   try {
     assert.throws(
-      () => executor.buildUrl("@cf/meta/llama-3.3-70b-instruct", true, 0, {}),
+      () => executor.buildUrl("@cf/meta/llama-3.3-70b-instruct-fp8-fast", true, 0, {}),
       /Account ID/
     );
   } finally {
@@ -76,10 +76,10 @@ test("CloudflareAIExecutor.buildHeaders uses API key or access token and stream
 test("CloudflareAIExecutor.transformRequest preserves plain-string content", () => {
   const executor = new CloudflareAIExecutor();
   const body = {
-    model: "@cf/meta/llama-3.3-70b-instruct",
+    model: "@cf/meta/llama-3.3-70b-instruct-fp8-fast",
     messages: [{ role: "user", content: "hi" }],
   };
-  const out = executor.transformRequest("@cf/meta/llama-3.3-70b-instruct", body, true, {} as any);
+  const out = executor.transformRequest("@cf/meta/llama-3.3-70b-instruct-fp8-fast", body, true, {} as any);
   assert.deepEqual((out as any).messages, [{ role: "user", content: "hi" }]);
 });
 
@@ -88,7 +88,7 @@ test("CloudflareAIExecutor.transformRequest preserves plain-string content", ()
 test("CloudflareAIExecutor.transformRequest flattens content-part arrays to strings (#2539)", () => {
   const executor = new CloudflareAIExecutor();
   const body = {
-    model: "@cf/meta/llama-3.3-70b-instruct",
+    model: "@cf/meta/llama-3.3-70b-instruct-fp8-fast",
     messages: [
       {
         role: "user",
@@ -101,7 +101,7 @@ test("CloudflareAIExecutor.transformRequest flattens content-part arrays to stri
     ],
   };
   const out = executor.transformRequest(
-    "@cf/meta/llama-3.3-70b-instruct",
+    "@cf/meta/llama-3.3-70b-instruct-fp8-fast",
     body,
     false,
     {} as any
@@ -132,11 +132,11 @@ test("CloudflareAIExecutor.execute uses inherited BaseExecutor flow successfully
 
   try {
     const body = {
-      model: "@cf/meta/llama-3.3-70b-instruct",
+      model: "@cf/meta/llama-3.3-70b-instruct-fp8-fast",
       messages: [{ role: "user", content: "hello" }],
     };
     const result = await executor.execute({
-      model: "@cf/meta/llama-3.3-70b-instruct",
+      model: "@cf/meta/llama-3.3-70b-instruct-fp8-fast",
       body,
       stream: false,
       credentials: {