Tests: align GPT-5.4 fallback pricing expectations

src/agents/context.lookup.test.ts

@@ -34,6 +34,38 @@ describe("lookupContextTokens", () => {
     expect(lookupContextTokens("openrouter/claude-sonnet")).toBe(321_000);
   });
 
+  it("adds GPT-5.4 context windows from forward-compat discovery fallbacks", async () => {
+    vi.doMock("../config/config.js", () => ({
+      loadConfig: () => ({
+        models: { providers: {} },
+      }),
+    }));
+    vi.doMock("./models-config.js", () => ({
+      ensureOpenClawModelsJson: vi.fn(async () => {}),
+    }));
+    vi.doMock("./agent-paths.js", () => ({
+      resolveOpenClawAgentDir: () => "/tmp/openclaw-agent",
+    }));
+    vi.doMock("./pi-model-discovery.js", () => ({
+      discoverAuthStorage: vi.fn(() => ({})),
+      discoverModels: vi.fn(() => ({
+        getAll: () => [
+          {
+            provider: "openai",
+            id: "gpt-5.2",
+            contextWindow: 400_000,
+          },
+        ],
+      })),
+    }));
+
+    const { lookupContextTokens } = await import("./context.js");
+    await vi.waitFor(() => {
+      expect(lookupContextTokens("gpt-5.4")).toBe(1_050_000);
+      expect(lookupContextTokens("gpt-5.4-pro")).toBe(1_050_000);
+    });
+  });
+
   it("does not skip eager warmup when --profile is followed by -- terminator", async () => {
     const loadConfigMock = vi.fn(() => ({ models: {} }));
     vi.doMock("../config/config.js", () => ({

src/agents/context.ts

@@ -1,17 +1,19 @@
 // Lazy-load pi-coding-agent model metadata so we can infer context windows when
 // the agent reports a model id. This includes custom models.json entries.
 
+import type { Api, Model } from "@mariozechner/pi-ai";
 import { loadConfig } from "../config/config.js";
 import type { OpenClawConfig } from "../config/config.js";
 import { computeBackoff, type BackoffPolicy } from "../infra/backoff.js";
 import { consumeRootOptionToken, FLAG_TERMINATOR } from "../infra/cli-root-options.js";
 import { resolveOpenClawAgentDir } from "./agent-paths.js";
+import { augmentKnownForwardCompatModels } from "./model-forward-compat.js";
 import { ensureOpenClawModelsJson } from "./models-config.js";
 
-type ModelEntry = { id: string; contextWindow?: number };
+type ModelEntry = { id: string; provider?: string; contextWindow?: number };
 type ModelRegistryLike = {
-  getAvailable?: () => ModelEntry[];
-  getAll: () => ModelEntry[];
+  getAvailable?: () => Model<Api>[];
+  getAll: () => Model<Api>[];
 };
 type ConfigModelEntry = { id?: string; contextWindow?: number };
 type ProviderConfigEntry = { models?: ConfigModelEntry[] };
@@ -156,10 +158,11 @@ function ensureContextWindowCacheLoaded(): Promise<void> {
       const agentDir = resolveOpenClawAgentDir();
       const authStorage = discoverAuthStorage(agentDir);
       const modelRegistry = discoverModels(authStorage, agentDir) as unknown as ModelRegistryLike;
-      const models =
+      const models = augmentKnownForwardCompatModels(
         typeof modelRegistry.getAvailable === "function"
           ? modelRegistry.getAvailable()
-          : modelRegistry.getAll();
+          : modelRegistry.getAll(),
+      );
       applyDiscoveredContextWindows({
         cache: MODEL_CACHE,
         models,

src/agents/live-model-filter.test.ts

@@ -0,0 +1,13 @@
+import { describe, expect, it } from "vitest";
+import { isModernModelRef } from "./live-model-filter.js";
+
+describe("isModernModelRef", () => {
+  it("accepts new openai gpt-5.4 refs", () => {
+    expect(isModernModelRef({ provider: "openai", id: "gpt-5.4" })).toBe(true);
+    expect(isModernModelRef({ provider: "openai", id: "gpt-5.4-pro" })).toBe(true);
+  });
+
+  it("keeps rejecting older openai refs outside the allowlist", () => {
+    expect(isModernModelRef({ provider: "openai", id: "gpt-4.1" })).toBe(false);
+  });
+});

src/agents/live-model-filter.ts

@@ -46,6 +46,8 @@ export function isModernModelRef(ref: ModelRef): boolean {
   }
 
   if (provider === "openai") {
+    // Keep the broader prefix match for GPT-5.x families so live tests keep opting into
+    // fresh OpenAI minor variants before the forward-compat catalog learns each exact ID.
     return matchesExactOrPrefix(id, OPENAI_MODELS);
   }
 

src/agents/model-compat.test.ts

@@ -307,7 +307,7 @@ describe("resolveForwardCompatModel", () => {
     expect(model?.reasoning).toBe(true);
     expect(model?.contextWindow).toBe(1_050_000);
     expect(model?.maxTokens).toBe(128_000);
-    expect(model?.cost).toEqual({ input: 0, output: 0, cacheRead: 0, cacheWrite: 0 });
+    expect(model?.cost).toEqual({ input: 2.5, output: 15, cacheRead: 0.25, cacheWrite: 0 });
   });
 
   it("resolves openai gpt-5.4-pro via template fallback", () => {
@@ -320,6 +320,7 @@ describe("resolveForwardCompatModel", () => {
     expect(model?.baseUrl).toBe("https://api.openai.com/v1");
     expect(model?.contextWindow).toBe(1_050_000);
     expect(model?.maxTokens).toBe(128_000);
+    expect(model?.cost).toEqual({ input: 30, output: 180, cacheRead: 0, cacheWrite: 0 });
   });
 
   it("resolves openai-codex gpt-5.4 via codex template fallback", () => {

src/agents/model-forward-compat.ts

@@ -10,10 +10,14 @@ const OPENAI_GPT_54_CONTEXT_TOKENS = 1_050_000;
 const OPENAI_GPT_54_MAX_TOKENS = 128_000;
 const OPENAI_GPT_54_TEMPLATE_MODEL_IDS = ["gpt-5.2"] as const;
 const OPENAI_GPT_54_PRO_TEMPLATE_MODEL_IDS = ["gpt-5.2-pro", "gpt-5.2"] as const;
+const OPENAI_GPT_54_COST = { input: 2.5, output: 15, cacheRead: 0.25, cacheWrite: 0 } as const;
+// OpenAI currently publishes no cached-input price for GPT-5.4 Pro.
+const OPENAI_GPT_54_PRO_COST = { input: 30, output: 180, cacheRead: 0, cacheWrite: 0 } as const;
 
 const OPENAI_CODEX_GPT_54_MODEL_ID = "gpt-5.4";
 const OPENAI_CODEX_GPT_54_TEMPLATE_MODEL_IDS = ["gpt-5.3-codex", "gpt-5.2-codex"] as const;
 const OPENAI_CODEX_GPT_53_MODEL_ID = "gpt-5.3-codex";
+const OPENAI_CODEX_GPT_53_SPARK_MODEL_ID = "gpt-5.3-codex-spark";
 const OPENAI_CODEX_TEMPLATE_MODEL_IDS = ["gpt-5.2-codex"] as const;
 
 const ANTHROPIC_OPUS_46_MODEL_ID = "claude-opus-4-6";
@@ -55,35 +59,23 @@ function resolveOpenAIGpt54ForwardCompatModel(
     return undefined;
   }
 
-  return (
-    cloneFirstTemplateModel({
-      normalizedProvider,
-      trimmedModelId,
-      templateIds: [...templateIds],
-      modelRegistry,
-      patch: {
-        api: "openai-responses",
-        provider: normalizedProvider,
-        baseUrl: "https://api.openai.com/v1",
-        reasoning: true,
-        input: ["text", "image"],
-        contextWindow: OPENAI_GPT_54_CONTEXT_TOKENS,
-        maxTokens: OPENAI_GPT_54_MAX_TOKENS,
-      },
-    }) ??
-    normalizeModelCompat({
-      id: trimmedModelId,
-      name: trimmedModelId,
+  const template = cloneFirstTemplateModel({
+    normalizedProvider,
+    trimmedModelId,
+    templateIds: [...templateIds],
+    modelRegistry,
+    patch: {
       api: "openai-responses",
       provider: normalizedProvider,
       baseUrl: "https://api.openai.com/v1",
       reasoning: true,
       input: ["text", "image"],
-      cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
       contextWindow: OPENAI_GPT_54_CONTEXT_TOKENS,
       maxTokens: OPENAI_GPT_54_MAX_TOKENS,
-    } as Model<Api>)
-  );
+    },
+  });
+
+  return buildOpenAIGpt54FallbackModel(trimmedModelId, template);
 }
 
 function cloneFirstTemplateModel(params: {
@@ -109,6 +101,179 @@ function cloneFirstTemplateModel(params: {
   return undefined;
 }
 
+function cloneSyntheticTemplateModel(params: {
+  models: Model<Api>[];
+  normalizedProvider: string;
+  trimmedModelId: string;
+  templateIds: readonly string[];
+  patch?: Partial<Model<Api>>;
+}): Model<Api> | undefined {
+  const { models, normalizedProvider, trimmedModelId, templateIds, patch } = params;
+  for (const templateId of [...new Set(templateIds)].filter(Boolean)) {
+    const template =
+      models.find(
+        (model) =>
+          normalizeProviderId(model.provider) === normalizedProvider &&
+          model.id.trim().toLowerCase() === templateId.toLowerCase(),
+      ) ?? null;
+    if (!template) {
+      continue;
+    }
+    return normalizeModelCompat({
+      ...template,
+      id: trimmedModelId,
+      name: trimmedModelId,
+      ...patch,
+    } as Model<Api>);
+  }
+  return undefined;
+}
+
+function buildOpenAIGpt54FallbackModel(modelId: string, template?: Model<Api> | null): Model<Api> {
+  return normalizeModelCompat({
+    ...template,
+    id: modelId,
+    name: modelId,
+    api: "openai-responses",
+    provider: "openai",
+    baseUrl: "https://api.openai.com/v1",
+    reasoning: true,
+    input: ["text", "image"],
+    cost:
+      modelId.toLowerCase() === OPENAI_GPT_54_PRO_MODEL_ID
+        ? OPENAI_GPT_54_PRO_COST
+        : OPENAI_GPT_54_COST,
+    contextWindow: OPENAI_GPT_54_CONTEXT_TOKENS,
+    maxTokens: OPENAI_GPT_54_MAX_TOKENS,
+  } as Model<Api>);
+}
+
+function buildOpenAICodexSparkFallbackModel(template?: Model<Api> | null): Model<Api> {
+  return normalizeModelCompat({
+    ...template,
+    id: OPENAI_CODEX_GPT_53_SPARK_MODEL_ID,
+    name: OPENAI_CODEX_GPT_53_SPARK_MODEL_ID,
+    api: "openai-codex-responses",
+    provider: "openai-codex",
+    baseUrl: "https://chatgpt.com/backend-api",
+    reasoning: true,
+    input: ["text", "image"],
+    cost: template?.cost ?? { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
+    contextWindow: template?.contextWindow ?? DEFAULT_CONTEXT_TOKENS,
+    maxTokens: template?.maxTokens ?? DEFAULT_CONTEXT_TOKENS,
+  } as Model<Api>);
+}
+
+export function augmentKnownForwardCompatModels(models: Model<Api>[]): Model<Api>[] {
+  const next = [...models];
+  const existing = new Set(
+    next.map((model) => `${normalizeProviderId(model.provider)}::${model.id.trim().toLowerCase()}`),
+  );
+  const hasProvider = (provider: string) =>
+    next.some((model) => normalizeProviderId(model.provider) === provider);
+  const pushIfMissing = (provider: string, id: string, model: Model<Api> | undefined) => {
+    const key = `${normalizeProviderId(provider)}::${id.trim().toLowerCase()}`;
+    if (existing.has(key) || !model) {
+      return;
+    }
+    next.push(model);
+    existing.add(key);
+  };
+
+  if (hasProvider("openai")) {
+    pushIfMissing(
+      "openai",
+      OPENAI_GPT_54_MODEL_ID,
+      buildOpenAIGpt54FallbackModel(
+        OPENAI_GPT_54_MODEL_ID,
+        cloneSyntheticTemplateModel({
+          models: next,
+          normalizedProvider: "openai",
+          trimmedModelId: OPENAI_GPT_54_MODEL_ID,
+          templateIds: OPENAI_GPT_54_TEMPLATE_MODEL_IDS,
+          patch: {
+            api: "openai-responses",
+            provider: "openai",
+            baseUrl: "https://api.openai.com/v1",
+            reasoning: true,
+            input: ["text", "image"],
+            contextWindow: OPENAI_GPT_54_CONTEXT_TOKENS,
+            maxTokens: OPENAI_GPT_54_MAX_TOKENS,
+          },
+        }),
+      ),
+    );
+    pushIfMissing(
+      "openai",
+      OPENAI_GPT_54_PRO_MODEL_ID,
+      buildOpenAIGpt54FallbackModel(
+        OPENAI_GPT_54_PRO_MODEL_ID,
+        cloneSyntheticTemplateModel({
+          models: next,
+          normalizedProvider: "openai",
+          trimmedModelId: OPENAI_GPT_54_PRO_MODEL_ID,
+          templateIds: OPENAI_GPT_54_PRO_TEMPLATE_MODEL_IDS,
+          patch: {
+            api: "openai-responses",
+            provider: "openai",
+            baseUrl: "https://api.openai.com/v1",
+            reasoning: true,
+            input: ["text", "image"],
+            contextWindow: OPENAI_GPT_54_CONTEXT_TOKENS,
+            maxTokens: OPENAI_GPT_54_MAX_TOKENS,
+          },
+        }),
+      ),
+    );
+  }
+
+  if (hasProvider("openai-codex")) {
+    pushIfMissing(
+      "openai-codex",
+      OPENAI_CODEX_GPT_54_MODEL_ID,
+      cloneSyntheticTemplateModel({
+        models: next,
+        normalizedProvider: "openai-codex",
+        trimmedModelId: OPENAI_CODEX_GPT_54_MODEL_ID,
+        templateIds: OPENAI_CODEX_GPT_54_TEMPLATE_MODEL_IDS,
+      }) ??
+        normalizeModelCompat({
+          id: OPENAI_CODEX_GPT_54_MODEL_ID,
+          name: OPENAI_CODEX_GPT_54_MODEL_ID,
+          api: "openai-codex-responses",
+          provider: "openai-codex",
+          baseUrl: "https://chatgpt.com/backend-api",
+          reasoning: true,
+          input: ["text", "image"],
+          cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
+          contextWindow: DEFAULT_CONTEXT_TOKENS,
+          maxTokens: DEFAULT_CONTEXT_TOKENS,
+        } as Model<Api>),
+    );
+    pushIfMissing(
+      "openai-codex",
+      OPENAI_CODEX_GPT_53_SPARK_MODEL_ID,
+      buildOpenAICodexSparkFallbackModel(
+        cloneSyntheticTemplateModel({
+          models: next,
+          normalizedProvider: "openai-codex",
+          trimmedModelId: OPENAI_CODEX_GPT_53_SPARK_MODEL_ID,
+          templateIds: [OPENAI_CODEX_GPT_53_MODEL_ID, ...OPENAI_CODEX_TEMPLATE_MODEL_IDS],
+          patch: {
+            api: "openai-codex-responses",
+            provider: "openai-codex",
+            baseUrl: "https://chatgpt.com/backend-api",
+            reasoning: true,
+            input: ["text", "image"],
+          },
+        }),
+      ),
+    );
+  }
+
+  return next;
+}
+
 const CODEX_GPT54_ELIGIBLE_PROVIDERS = new Set(["openai-codex"]);
 const CODEX_GPT53_ELIGIBLE_PROVIDERS = new Set(["openai-codex", "github-copilot"]);
 

src/agents/pi-embedded-runner/model.test.ts

@@ -251,7 +251,6 @@ describe("resolveModel", () => {
     expect(result.model?.contextWindow).toBe(262144);
     expect(result.model?.maxTokens).toBe(32768);
   });
-
   it("propagates reasoning from matching configured fallback model", () => {
     const cfg = {
       models: {
@@ -446,6 +445,30 @@ describe("resolveModel", () => {
     });
   });
 
+  it("uses GPT-5.4 Pro pricing when cloning an older openai template", () => {
+    mockDiscoveredModel({
+      provider: "openai",
+      modelId: "gpt-5.2",
+      templateModel: buildForwardCompatTemplate({
+        id: "gpt-5.2",
+        name: "GPT-5.2",
+        provider: "openai",
+        api: "openai-responses",
+        baseUrl: "https://api.openai.com/v1",
+      }),
+    });
+
+    const result = resolveModel("openai", "gpt-5.4-pro", "/tmp/agent");
+
+    expect(result.error).toBeUndefined();
+    expect(result.model?.cost).toEqual({
+      input: 30,
+      output: 180,
+      cacheRead: 0,
+      cacheWrite: 0,
+    });
+  });
+
   it("builds an anthropic forward-compat fallback for claude-opus-4-6", () => {
     mockDiscoveredModel({
       provider: "anthropic",

src/auto-reply/thinking.test.ts

@@ -42,6 +42,11 @@ describe("normalizeThinkLevel", () => {
 });
 
 describe("listThinkingLevels", () => {
+  it("includes xhigh for openai gpt-5.4 refs", () => {
+    expect(listThinkingLevels("openai", "gpt-5.4")).toContain("xhigh");
+    expect(listThinkingLevels("openai", "gpt-5.4-pro")).toContain("xhigh");
+  });
+
   it("includes xhigh for codex models", () => {
     expect(listThinkingLevels(undefined, "gpt-5.2-codex")).toContain("xhigh");
     expect(listThinkingLevels(undefined, "gpt-5.3-codex")).toContain("xhigh");

src/commands/models.list.e2e.test.ts

@@ -264,6 +264,42 @@ describe("models list/status", () => {
     expect(payload.models[0]?.available).toBe(false);
   });
 
+  it("loadModelRegistry augments raw OpenAI models with GPT-5.4 fallbacks", async () => {
+    modelRegistryState.models = [
+      {
+        provider: "openai",
+        id: "gpt-5.2",
+        name: "GPT-5.2",
+        api: "openai-responses",
+        input: ["text", "image"],
+        baseUrl: "https://api.openai.com/v1",
+        contextWindow: 400_000,
+        maxTokens: 128_000,
+        reasoning: true,
+        cost: { input: 1.75, output: 14, cacheRead: 0.175, cacheWrite: 0 },
+      },
+    ];
+    modelRegistryState.available = modelRegistryState.models;
+
+    const { models } = await loadModelRegistry({});
+    expect(models).toEqual(
+      expect.arrayContaining([
+        expect.objectContaining({
+          provider: "openai",
+          id: "gpt-5.4",
+          api: "openai-responses",
+          contextWindow: 1_050_000,
+        }),
+        expect.objectContaining({
+          provider: "openai",
+          id: "gpt-5.4-pro",
+          api: "openai-responses",
+          contextWindow: 1_050_000,
+        }),
+      ]),
+    );
+  });
+
   it("models list does not treat availability-unavailable code as discovery fallback", async () => {
     configureGoogleAntigravityModel("claude-opus-4-6-thinking");
     modelRegistryState.getAllError = Object.assign(new Error("model discovery failed"), {

src/commands/models/list.registry.ts

@@ -8,6 +8,7 @@ import {
   resolveAwsSdkEnvVarName,
   resolveEnvApiKey,
 } from "../../agents/model-auth.js";
+import { augmentKnownForwardCompatModels } from "../../agents/model-forward-compat.js";
 import { ensureOpenClawModelsJson } from "../../agents/models-config.js";
 import { discoverAuthStorage, discoverModels } from "../../agents/pi-model-discovery.js";
 import type { OpenClawConfig } from "../../config/config.js";
@@ -99,7 +100,7 @@ export async function loadModelRegistry(cfg: OpenClawConfig) {
   const agentDir = resolveOpenClawAgentDir();
   const authStorage = discoverAuthStorage(agentDir);
   const registry = discoverModels(authStorage, agentDir);
-  const models = registry.getAll();
+  const models = augmentKnownForwardCompatModels(registry.getAll());
   let availableKeys: Set<string> | undefined;
   let availabilityErrorMessage: string | undefined;