test(models): modernize leftover gpt-4.1 fixtures to gpt-5.6

evals/capability/README.md

@@ -138,14 +138,14 @@ bun run eval:capability -- --case simple-health --provider xai --model grok-4.5
 
 # Multi-model matrix (cases × variants)
 bun run eval:capability -- \
-  --matrix "xai:grok-4.5,openai:gpt-4.1" \
+  --matrix "xai:grok-4.5,openai:gpt-5.6" \
   --out evals/capability/results/matrix.json
 
 # Faster live matrix (independent cells; default is serial)
 bun run eval:capability -- --provider <name> --model <id> --concurrency 4
 
 # Labeled variants
-bun run eval:capability -- --matrix "fast=xai:grok-4.5,strong=openai:gpt-4.1"
+bun run eval:capability -- --matrix "fast=xai:grok-4.5,strong=openai:gpt-5.6"
 
 # Baseline improve/regress (keys by variantId::caseId)
 bun run eval:capability -- --provider <name> --model <id> \

evals/capability/lib.test.ts

@@ -314,17 +314,17 @@ describe("parseMatrix", () => {
   });
 
   test("parses provider:model cells", () => {
-    const v = parseMatrix("xai:grok-4.5,openai:gpt-4.1", {});
+    const v = parseMatrix("xai:grok-4.5,openai:gpt-5.6", {});
     expect(v).toHaveLength(2);
     expect(v[0]).toEqual({
       id: "xai:grok-4.5",
       provider: "xai",
       model: "grok-4.5",
     });
     expect(v[1]).toEqual({
-      id: "openai:gpt-4.1",
+      id: "openai:gpt-5.6",
       provider: "openai",
-      model: "gpt-4.1",
+      model: "gpt-5.6",
     });
   });
 
@@ -693,7 +693,7 @@ describe("detectProviderFallback", () => {
     const info = detectProviderFallback({
       requestedProvider: "xai",
       resolvedProvider: "openai",
-      resolvedModel: "gpt-4.1",
+      resolvedModel: "gpt-5.6",
     });
     expect(info?.requestedProvider).toBe("xai");
     expect(info?.resolvedProvider).toBe("openai");
@@ -704,12 +704,12 @@ describe("detectProviderFallback", () => {
       requestedProvider: "xai",
       requestedModel: "grok-4.5",
       resolvedProvider: "openai",
-      resolvedModel: "gpt-4.1",
+      resolvedModel: "gpt-5.6",
     });
     expect(info).not.toBeNull();
     const message = formatProviderFallback(defined(info));
     expect(message).toContain("xai/grok-4.5");
-    expect(message).toContain("openai/gpt-4.1");
+    expect(message).toContain("openai/gpt-5.6");
   });
 });
 
@@ -776,10 +776,10 @@ describe("resolveRequestedProviderModel", () => {
 
   test("a matrix cell's own override wins over the run's resolved labels", () => {
     const requested = resolveRequestedProviderModel(
-      { provider: "openai", model: "gpt-4.1" },
+      { provider: "openai", model: "gpt-5.6" },
       { provider: "xai", model: "grok-4.5" },
     );
-    expect(requested).toEqual({ provider: "openai", model: "gpt-4.1" });
+    expect(requested).toEqual({ provider: "openai", model: "gpt-5.6" });
   });
 
   test("the '(default)' placeholder from an unresolvable ambient probe is not a real request", () => {

packages/first-class-providers/src/providers.test.ts

@@ -69,6 +69,8 @@ describe("FIRST_CLASS_PROVIDERS", () => {
     for (const model of api?.maxCompletionTokensModels ?? []) {
       expect(api?.models).toContain(model);
     }
+    // Keep: gpt-4.1 remains the catalog's non-reasoning OpenAI API model and
+    // must not join the max_completion_tokens set.
     expect(api?.maxCompletionTokensModels).not.toContain("gpt-4.1");
   });
 

packages/first-class-providers/src/providers.ts

@@ -19,6 +19,9 @@ const OPENAI_API_MODELS = [
   "gpt-6-astra",
   "gpt-5.4",
   "gpt-5.4-mini",
+  // Keep: still a first-party API catalog id; non-reasoning so it stays on
+  // max_tokens (not in OPENAI_API_MAX_COMPLETION_TOKENS_MODELS). Dropping it
+  // would change Connect/API-path model lists.
   "gpt-4.1",
   "o3",
   "o4-mini",

src/agent/prompt-sizes.ts

@@ -59,7 +59,7 @@ export const CANONICAL_PROMPT_ENV: EnvironmentInfo = {
 
 const GROK_PROVIDER = { providerName: "xai/default", model: "grok-4.6" };
 // Default-family probe: anthropic/claude-sonnet-4 hits the claude row and
-// openai/gpt-4.1 hits the gpt row, so an unrecognized provider is the probe
+// openai/gpt-5.6 hits the gpt row, so an unrecognized provider is the probe
 // that still resolves to the default family (no residual).
 const MUSE_PROVIDER = {
   providerName: "opencode-go",
@@ -73,7 +73,7 @@ const CLAUDE_PROVIDER = {
   providerName: "anthropic",
   model: "claude-sonnet-4",
 };
-const GPT_PROVIDER = { providerName: "openai", model: "gpt-4.1" };
+const GPT_PROVIDER = { providerName: "openai", model: "gpt-5.6" };
 
 /** Families in the size table: default (no residual), muse, grok, claude, gpt. */
 export type PromptSizeFamily = "default" | "muse" | "grok" | "claude" | "gpt";

src/agent/tool-schema-normalize.test.ts

@@ -122,7 +122,7 @@ describe("normalizeToolDefinitionsForProvider", () => {
   test("non-kimi providers get identity schemas (recursive present kept)", () => {
     for (const ctx of [
       { providerName: "anthropic", model: "claude-sonnet-4" },
-      { providerName: "openai", model: "gpt-4.1" },
+      { providerName: "openai", model: "gpt-5.6" },
       { providerName: "xai/default", model: "grok-4.5" },
       { providerName: "opencode-go", model: "gpt-5.1" },
     ] as const) {

src/config/inference-sources.test.ts

@@ -172,6 +172,8 @@ describe("OpenAI reasoning max_completion_tokens quirk (CL-7785)", () => {
     // max_tokens. Adding a preset model requires a decision here AND in the
     // preset's maxCompletionTokensModels — the union below fails loudly
     // otherwise instead of silently sending max_tokens.
+    // Keep: gpt-4.1 is the catalog's remaining non-reasoning preset. The
+    // union with maxCompletionTokensModels must equal OPENAI_API_MODELS.
     const explicitMaxTokensModels = new Set(["gpt-4.1"]);
     expect([...flagged, ...explicitMaxTokensModels].sort()).toEqual(
       [...new Set(presetModels)].sort(),

src/tui/model-catalog.test.ts

@@ -11,12 +11,12 @@ describe("buildModelCatalog", () => {
   test("maps provider array to id/label picker options", () => {
     const options = buildModelCatalog([
       { name: "xai", models: ["grok-4", "grok-3"], label: "xAI" },
-      { name: "openai", models: ["gpt-4.1"] },
+      { name: "openai", models: ["gpt-5.6"] },
     ]);
     expect(options).toEqual([
       { id: "xai:grok-4", label: "grok-4 * [xAI]" },
       { id: "xai:grok-3", label: "grok-3 * [xAI]" },
-      { id: "openai:gpt-4.1", label: "gpt-4.1 * [openai]" },
+      { id: "openai:gpt-5.6", label: "gpt-5.6 * [openai]" },
     ]);
   });
 

tests/unit/pricing-fetcher.test.ts

@@ -25,13 +25,13 @@ test("parseModelsDevContextWindows reads limit.context per model", () => {
     },
     openai: {
       models: {
-        "gpt-4.1": { id: "openai/gpt-4.1", limit: { context: 1_000_000 } },
+        "gpt-5.6": { id: "openai/gpt-5.6", limit: { context: 1_000_000 } },
         "no-limit": { id: "openai/no-limit" },
       },
     },
   });
   expect(windows["z-ai/glm-4.6"]).toBe(64_000);
-  expect(windows["openai/gpt-4.1"]).toBe(1_000_000);
+  expect(windows["openai/gpt-5.6"]).toBe(1_000_000);
   expect(windows["openai/no-limit"]).toBeUndefined();
 });
 
@@ -53,8 +53,8 @@ test("parseModelsDevPricing supports multiple model IDs", () => {
   const models = parseModelsDevPricing({
     openai: {
       models: {
-        "gpt-4.1": {
-          id: "openai/gpt-4.1",
+        "gpt-5.6": {
+          id: "openai/gpt-5.6",
           input_cost_per_million: 2,
           output_cost_per_million: 8,
           cache_read_cost_per_million: 0.5,
@@ -72,7 +72,7 @@ test("parseModelsDevPricing supports multiple model IDs", () => {
     },
   });
 
-  expect(models["openai/gpt-4.1"]).toEqual({
+  expect(models["openai/gpt-5.6"]).toEqual({
     inputPricePerToken: 0.000002,
     outputPricePerToken: 0.000008,
     cacheReadPricePerToken: 0.0000005,