diff --git a/packages/ai/src/protocols/google-interactions.ts b/packages/ai/src/protocols/google-interactions.ts
index 51abf1bcea1c..36a8f877b64e 100644
--- a/packages/ai/src/protocols/google-interactions.ts
+++ b/packages/ai/src/protocols/google-interactions.ts
@@ -33,7 +33,7 @@ const Options = Schema.Struct({
   store: lenient(Schema.Boolean),
   thinkingLevel: lenient(ThinkingLevel),
   thinkingSummaries: lenient(knownString<"auto" | "none">()),
-  serviceTier: lenient(knownString<"standard" | "flex" | "priority">()),
+  serviceTier: lenient(knownString<"standard" | "flex" | "priority" | "deferred">()),
 })
 export type OptionsInput = typeof Options.Encoded
 export type ProviderOptionsInput = OptionsInput
diff --git a/packages/ai/src/protocols/utils/openai-options.ts b/packages/ai/src/protocols/utils/openai-options.ts
index f3e56be4a061..00f6592006ed 100644
--- a/packages/ai/src/protocols/utils/openai-options.ts
+++ b/packages/ai/src/protocols/utils/openai-options.ts
@@ -9,7 +9,8 @@ export type OpenAITextVerbosity = OpenResponsesOptions.TextVerbosity
 // in lockstep with `openai-node/src/resources/responses/responses.ts`.
 export const OpenAIResponseIncludables = OpenResponsesOptions.ResponseIncludables
 export type OpenAIResponseIncludable = OpenResponsesOptions.ResponseIncludable
-export const OpenAIServiceTiers = [...OpenResponsesOptions.ServiceTiers, "scale"] as const
+// Mirrors OpenAI's `ServiceTier` union from the official SDK.
+export const OpenAIServiceTiers = [...OpenResponsesOptions.ServiceTiers, "scale", "fast", "ultrafast"] as const
 export type OpenAIServiceTier = (typeof OpenAIServiceTiers)[number] | (string & {})
 
 export const OpenAIReasoningEffort = OpenResponsesOptions.ReasoningEffort
diff --git a/packages/ai/test/provider-options/openai.types.ts b/packages/ai/test/provider-options/openai.types.ts
index e2746750f48c..ddda75470734 100644
--- a/packages/ai/test/provider-options/openai.types.ts
+++ b/packages/ai/test/provider-options/openai.types.ts
@@ -9,6 +9,7 @@ LLM.request({ model: selected, prompt: "Hello", providerOptions: { reasoningEffo
 LLM.request({ model: selected, prompt: "Hello", providerOptions: { textVerbosity: "low" } })
 LLM.request({ model: selected, prompt: "Hello", providerOptions: { textVerbosity: "verbose" } })
 LLM.request({ model: selected, prompt: "Hello", providerOptions: { serviceTier: "scale" } })
+LLM.request({ model: selected, prompt: "Hello", providerOptions: { serviceTier: "ultrafast" } })
 LLM.request({ model: selected, prompt: "Hello", providerOptions: { serviceTier: "future-tier" } })
 LLM.request({ model: chat, prompt: "Hello", providerOptions: { reasoningEffort: "max" } })
 LLM.request({ model: chat, prompt: "Hello", providerOptions: { reasoningEffort: "experimental" } })
diff --git a/packages/ai/test/provider/openai-responses.test.ts b/packages/ai/test/provider/openai-responses.test.ts
index ca613295a321..e2ef36e446a6 100644
--- a/packages/ai/test/provider/openai-responses.test.ts
+++ b/packages/ai/test/provider/openai-responses.test.ts
@@ -325,11 +325,13 @@ describe("OpenAI Responses route", () => {
     }),
   )
 
-  it.effect("passes through provider-defined service tiers", () =>
+  it.effect("passes through provider-defined and future service tiers", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(LLMRequest.update(request, { providerOptions: { serviceTier: "scale" } }))
+      for (const serviceTier of ["scale", "ultrafast", "future-tier"]) {
+        const prepared = yield* compileRequest(LLMRequest.update(request, { providerOptions: { serviceTier } }))
 
-      expect(prepared.body.service_tier).toBe("scale")
+        expect(prepared.body.service_tier).toBe(serviceTier)
+      }
     }),
   )