bench: update codex models

Author: ammar-agent

.github/workflows/nightly-terminal-bench.yml

@@ -23,7 +23,7 @@ jobs:
         id: set-models
         run: |
           if [ "${{ inputs.models }}" = "all" ] || [ -z "${{ inputs.models }}" ]; then
-            echo 'models=["anthropic:claude-sonnet-4-5","openai:gpt-5-codex"]' >> $GITHUB_OUTPUT
+            echo 'models=["anthropic:claude-sonnet-4-5","openai:gpt-5.1-codex"]' >> $GITHUB_OUTPUT
           else
             # Convert comma-separated to JSON array
             models="${{ inputs.models }}"

.github/workflows/terminal-bench.yml

@@ -61,7 +61,7 @@ on:
         required: false
         type: string
       model_name:
-        description: 'Model to use (e.g., anthropic:claude-sonnet-4-5, openai:gpt-5-codex)'
+        description: 'Model to use (e.g., anthropic:claude-sonnet-4-5, openai:gpt-5.1-codex)'
         required: false
         type: string
       thinking_level:

src/constants/knownModels.ts

@@ -66,13 +66,18 @@ const MODEL_DEFINITIONS = {
   GPT_MINI: {
     provider: "openai",
     providerModelId: "gpt-5.1-codex-mini",
+    aliases: ["codex-mini"],
   },
 } as const satisfies Record<string, KnownModelDefinition>;
 
 export type KnownModelKey = keyof typeof MODEL_DEFINITIONS;
+const MODEL_DEFINITION_ENTRIES = Object.entries(MODEL_DEFINITIONS) as Array<
+  [KnownModelKey, KnownModelDefinition]
+>;
+
 
 export const KNOWN_MODELS = Object.fromEntries(
-  Object.entries(MODEL_DEFINITIONS).map(([key, definition]) => [
+  MODEL_DEFINITION_ENTRIES.map(([key, definition]) => [
     key,
     {
       ...definition,

tests/ipcMain/anthropic1MContext.test.ts

@@ -4,6 +4,7 @@ import {
   createEventCollector,
   assertStreamSuccess,
   buildLargeHistory,
+  modelString,
 } from "./helpers";
 
 // Skip all tests if TEST_INTEGRATION is not set
@@ -42,8 +43,7 @@ describeIntegration("IpcMain anthropic 1M context integration tests", () => {
           env.mockIpcRenderer,
           workspaceId,
           "Summarize the context above in one word.",
-          "anthropic",
-          "claude-sonnet-4-5",
+          modelString("anthropic", "claude-sonnet-4-5"),
           {
             providerOptions: {
               anthropic: {
@@ -76,8 +76,7 @@ describeIntegration("IpcMain anthropic 1M context integration tests", () => {
           env.mockIpcRenderer,
           workspaceId,
           "Summarize the context above in one word.",
-          "anthropic",
-          "claude-sonnet-4-5",
+          modelString("anthropic", "claude-sonnet-4-5"),
           {
             providerOptions: {
               anthropic: {

tests/ipcMain/forkWorkspace.test.ts

@@ -13,6 +13,7 @@ import {
   createEventCollector,
   assertStreamSuccess,
   waitFor,
+  modelString,
 } from "./helpers";
 import { detectDefaultTrunkBranch } from "../../src/git";
 import { HistoryService } from "../../src/services/historyService";
@@ -100,9 +101,8 @@ describeIntegration("IpcMain fork workspace integration tests", () => {
           env.mockIpcRenderer,
           forkedWorkspaceId,
           "What is 2+2? Answer with just the number.",
-          "anthropic",
-          "claude-sonnet-4-5"
-        );
+          modelString("anthropic", "claude-sonnet-4-5")
+          );
         expect(sendResult.success).toBe(true);
 
         // Verify stream completes successfully
@@ -154,9 +154,8 @@ describeIntegration("IpcMain fork workspace integration tests", () => {
           env.mockIpcRenderer,
           forkedWorkspaceId,
           "What word did I ask you to remember? Reply with just the word.",
-          "anthropic",
-          "claude-sonnet-4-5"
-        );
+          modelString("anthropic", "claude-sonnet-4-5")
+          );
         expect(sendResult.success).toBe(true);
 
         // Verify stream completes successfully
@@ -206,16 +205,12 @@ describeIntegration("IpcMain fork workspace integration tests", () => {
             env.mockIpcRenderer,
             sourceWorkspaceId,
             "What is 5+5? Answer with just the number.",
-            "anthropic",
-            "claude-sonnet-4-5"
-          ),
+            modelString("anthropic", "claude-sonnet-4-5")),
           sendMessageWithModel(
             env.mockIpcRenderer,
             forkedWorkspaceId,
             "What is 3+3? Answer with just the number.",
-            "anthropic",
-            "claude-sonnet-4-5"
-          ),
+            modelString("anthropic", "claude-sonnet-4-5")),
         ]);
 
         expect(sourceResult.success).toBe(true);
@@ -253,9 +248,8 @@ describeIntegration("IpcMain fork workspace integration tests", () => {
           env.mockIpcRenderer,
           sourceWorkspaceId,
           "Count from 1 to 10, one number per line. Then say 'Done counting.'",
-          "anthropic",
-          "claude-sonnet-4-5"
-        );
+          modelString("anthropic", "claude-sonnet-4-5")
+          );
 
         // Wait for stream to start and produce some content
         const sourceCollector = createEventCollector(env.sentEvents, sourceWorkspaceId);
@@ -286,9 +280,8 @@ describeIntegration("IpcMain fork workspace integration tests", () => {
           env.mockIpcRenderer,
           forkedWorkspaceId,
           "What is 7+3? Answer with just the number.",
-          "anthropic",
-          "claude-sonnet-4-5"
-        );
+          modelString("anthropic", "claude-sonnet-4-5")
+          );
         expect(forkedSendResult.success).toBe(true);
 
         // Verify forked workspace stream completes successfully

tests/ipcMain/helpers.ts

@@ -14,6 +14,7 @@ import * as os from "os";
 import { detectDefaultTrunkBranch } from "../../src/git";
 import type { TestEnvironment } from "./setup";
 import type { RuntimeConfig } from "../../src/types/runtime";
+import { KNOWN_MODELS } from "../../src/constants/knownModels";
 import type { ToolPolicy } from "../../src/utils/tools/toolPolicy";
 
 // Test constants - centralized for consistency across all tests
@@ -46,6 +47,13 @@ export function modelString(provider: string, model: string): string {
 /**
  * Send a message via IPC
  */
+type SendMessageWithModelOptions = Omit<SendMessageOptions, "model"> & {
+  imageParts?: Array<{ url: string; mediaType: string }>;
+};
+
+const DEFAULT_MODEL_ID = KNOWN_MODELS.SONNET.id;
+const DEFAULT_PROVIDER = KNOWN_MODELS.SONNET.provider;
+
 export async function sendMessage(
   mockIpcRenderer: IpcRenderer,
   workspaceId: string,
@@ -61,19 +69,22 @@ export async function sendMessage(
 }
 
 /**
- * Send a message with a provider and model (convenience wrapper)
+ * Send a message with an explicit model id (defaults to SONNET).
  */
 export async function sendMessageWithModel(
   mockIpcRenderer: IpcRenderer,
   workspaceId: string,
   message: string,
-  provider = "anthropic",
-  model = "claude-sonnet-4-5",
-  options?: Omit<SendMessageOptions, "model">
+  modelId: string = DEFAULT_MODEL_ID,
+  options?: SendMessageWithModelOptions
 ): Promise<Result<void, SendMessageError>> {
+  const resolvedModel = modelId.includes(":")
+    ? modelId
+    : modelString(DEFAULT_PROVIDER, modelId);
+
   return sendMessage(mockIpcRenderer, workspaceId, message, {
     ...options,
-    model: modelString(provider, model),
+    model: resolvedModel,
   });
 }
 

tests/ipcMain/modelNotFound.test.ts

@@ -1,5 +1,5 @@
 import { setupWorkspace, shouldRunIntegrationTests, validateApiKeys } from "./setup";
-import { sendMessageWithModel, createEventCollector, waitFor } from "./helpers";
+import { sendMessageWithModel, createEventCollector, waitFor, modelString } from "./helpers";
 import { IPC_CHANNELS } from "../../src/constants/ipc-constants";
 import type { Result } from "../../src/types/result";
 import type { SendMessageError } from "../../src/types/errors";
@@ -30,9 +30,8 @@ describeIntegration("IpcMain model_not_found error handling", () => {
           env.mockIpcRenderer,
           workspaceId,
           "Hello",
-          "anthropic",
-          "invalid-model-that-does-not-exist-xyz123"
-        );
+          modelString("anthropic", "invalid-model-that-does-not-exist-xyz123")
+          );
 
         // Collect events to verify error classification
         const collector = createEventCollector(env.sentEvents, workspaceId);
@@ -69,9 +68,8 @@ describeIntegration("IpcMain model_not_found error handling", () => {
           env.mockIpcRenderer,
           workspaceId,
           "Hello",
-          "openai",
-          "gpt-nonexistent-model-xyz123"
-        );
+          modelString("openai", "gpt-nonexistent-model-xyz123")
+          );
 
         // Collect events to verify error classification
         const collector = createEventCollector(env.sentEvents, workspaceId);

tests/ipcMain/ollama.test.ts

@@ -4,6 +4,7 @@ import {
   createEventCollector,
   assertStreamSuccess,
   extractTextFromEvents,
+  modelString,
 } from "./helpers";
 import { spawn } from "child_process";
 
@@ -105,9 +106,8 @@ describeOllama("IpcMain Ollama integration tests", () => {
         env.mockIpcRenderer,
         workspaceId,
         "Say 'hello' and nothing else",
-        "ollama",
-        OLLAMA_MODEL
-      );
+        modelString("ollama", OLLAMA_MODEL)
+        );
 
       // Verify the IPC call succeeded
       expect(result.success).toBe(true);
@@ -139,9 +139,8 @@ describeOllama("IpcMain Ollama integration tests", () => {
         env.mockIpcRenderer,
         workspaceId,
         "What is the current date and time? Use the bash tool to find out.",
-        "ollama",
-        OLLAMA_MODEL
-      );
+        modelString("ollama", OLLAMA_MODEL)
+        );
 
       expect(result.success).toBe(true);
 
@@ -178,9 +177,8 @@ describeOllama("IpcMain Ollama integration tests", () => {
         env.mockIpcRenderer,
         workspaceId,
         "Read the README.md file and tell me what the first heading says.",
-        "ollama",
-        OLLAMA_MODEL
-      );
+        modelString("ollama", OLLAMA_MODEL)
+        );
 
       expect(result.success).toBe(true);
 
@@ -216,8 +214,7 @@ describeOllama("IpcMain Ollama integration tests", () => {
         env.mockIpcRenderer,
         workspaceId,
         "This should fail",
-        "ollama",
-        OLLAMA_MODEL,
+        modelString("ollama", OLLAMA_MODEL),
         {
           providerOptions: {
             ollama: {},

tests/ipcMain/openai-web-search.test.ts

@@ -1,5 +1,5 @@
 import { setupWorkspace, shouldRunIntegrationTests, validateApiKeys } from "./setup";
-import { sendMessageWithModel, createEventCollector, assertStreamSuccess } from "./helpers";
+import { sendMessageWithModel, createEventCollector, assertStreamSuccess, modelString } from "./helpers";
 
 // Skip all tests if TEST_INTEGRATION is not set
 const describeIntegration = shouldRunIntegrationTests() ? describe : describe.skip;
@@ -32,8 +32,7 @@ describeIntegration("OpenAI web_search integration tests", () => {
           workspaceId,
           "Use web search to find the current weather in San Francisco. " +
             "Then tell me if it's a good day for a picnic.",
-          "openai",
-          "gpt-5-codex",
+          modelString("openai", "gpt-5.1-codex-mini"),
           {
             thinkingLevel: "medium", // Ensure reasoning without excessive deliberation
           }

tests/ipcMain/resumeStream.test.ts

@@ -1,5 +1,5 @@
 import { setupWorkspace, shouldRunIntegrationTests, validateApiKeys } from "./setup";
-import { sendMessageWithModel, createEventCollector, waitFor } from "./helpers";
+import { sendMessageWithModel, createEventCollector, waitFor, modelString } from "./helpers";
 import { IPC_CHANNELS } from "../../src/constants/ipc-constants";
 import type { Result } from "../../src/types/result";
 import type { SendMessageError } from "../../src/types/errors";
@@ -31,9 +31,8 @@ describeIntegration("IpcMain resumeStream integration tests", () => {
           env.mockIpcRenderer,
           workspaceId,
           `Run this bash command: for i in 1 2 3; do sleep 0.5; done && echo '${expectedWord}'`,
-          "anthropic",
-          "claude-sonnet-4-5"
-        );
+          modelString("anthropic", "claude-sonnet-4-5")
+          );
 
         // Wait for stream to start
         const collector1 = createEventCollector(env.sentEvents, workspaceId);