@oh-my-pi/pi-catalog 16.3.13 → 16.3.15

package/CHANGELOG.md

@@ -2,6 +2,32 @@
 
 ## [Unreleased]
 
+## [16.3.15] - 2026-07-09
+
+### Added
+
+- Added support for Grok 4.5 model
+- Added `gpt-5.6` base models and `gpt-5.6-{luna,sol,terra}-pro` variants
+- Added `meta/muse-spark-1.1` model support
+- Added support for thinking modes on `poolside/laguna` models
+- Added generated GPT-5.6 Pro aliases (`gpt-5.6-{luna,sol,terra}-pro`) on the `openai` and `openai-codex` providers: each alias sends the base model id on the wire (`requestModelId`) with the new `reasoningMode: "pro"` marker, and re-derives from the current base rows on every catalog regeneration.
+
+### Changed
+
+- Updated cache read costs for Grok models
+- Reduced max token limit for Grok 4.3 model
+- Enabled prompt cache affinity for Grok models via the x-grok-conv-id header in OpenAI compatible endpoints
+- Enabled prompt cache affinity for Grok models via the x-grok-conv-id header
+- Marked direct xAI Grok Chat Completions models for `x-grok-conv-id` prompt-cache affinity.
+
+## [16.3.14] - 2026-07-09
+
+### Added
+
+- Added support for GPT-5.6 (Luna, Sol, Terra) model variants
+- Enabled expanded five-tier reasoning effort scale (minimal to xhigh) for GPT-5.6 models
+- Added GPT-5.6 (Terra/Luna/Sol) support for the new `max` reasoning tier: on wire-effort APIs (OpenAI Responses, Codex, Azure, openai-compat/OpenRouter models that advertise reasoning) user efforts shift up one notch — `xhigh` sends `max`, `high` sends `xhigh` — mirroring the Claude Fable/Opus 4.7+ five-tier mapping, and the exposed ladder becomes `minimal..xhigh` with `minimal` reaching the native `low` tier. Devin's per-tier GPT-5.6 sibling rows now collapse into `gpt-5-6-{luna,sol,terra}` logical models with the same shifted routing (`xhigh` → `-max`), plus `-fast` families that keep the direct `low..xhigh` `-priority` scale since Devin serves no `-max-priority` tier.
+
 ## [16.3.13] - 2026-07-09
 
 ### Added

package/dist/types/model-thinking.d.ts

@@ -14,12 +14,14 @@ import type { Api, CompatOf, Model, ModelSpec, ThinkingConfig } from "./types.js
  */
 type ApiModel<TApi extends Api = Api> = ModelSpec<TApi> | Model<TApi>;
 /**
- * Effort → wire-value map for the 5-tier adaptive scale (Opus 4.7+ and
- * Fable/Mythos 5 on the Messages API). User-facing efforts shift up one notch
- * so the top tier reaches the genuine "max" and "high" lands on Anthropic's
- * recommended "xhigh" coding/agentic default.
+ * Effort → wire-value map for a shifted five-tier scale (`low..max`):
+ * user-facing efforts shift up one notch so the top tier reaches the genuine
+ * "max" and "high" lands on the recommended "xhigh" coding/agentic default.
+ * Used by Anthropic adaptive models with a real xhigh tier (Opus 4.7+ and
+ * Fable/Mythos 5 on the Messages API) and by GPT-5.6+ wire-effort models,
+ * which expose the same genuine `max` tier above `xhigh`.
  */
-export declare const ANTHROPIC_ADAPTIVE_EFFORT_MAP_5_TIER: Readonly<Partial<Record<Effort, string>>>;
+export declare const SHIFTED_FIVE_TIER_EFFORT_MAP: Readonly<Partial<Record<Effort, string>>>;
 /**
  * Effort → wire-value map for the legacy 4-tier adaptive scale (Opus 4.6,
  * Sonnet 4.6+, and every adaptive model on Bedrock Converse). `low..high` pass

package/dist/types/provider-models/openai-compat.d.ts

@@ -54,6 +54,18 @@ export interface OpenAIModelManagerConfig {
     fetch?: FetchImpl;
 }
 export declare function openaiModelManagerOptions(config?: OpenAIModelManagerConfig): ModelManagerOptions<"openai-responses">;
+/**
+ * Re-derive the generated pro-reasoning aliases (`gpt-5.6-*-pro`) for the
+ * first-party `openai`/`openai-codex` gpt-5.6 rows. Each alias inherits the
+ * base row's metadata, requests the base wire id via `requestModelId`, and
+ * sets `reasoningMode: "pro"` so Responses-family request builders emit
+ * `reasoning: { mode: "pro" }`. Called by the models.json generator after all
+ * sources merge: stale copies of the owned aliases (previous snapshot) are
+ * dropped and re-projected from the current base rows so alias metadata always
+ * tracks the base, while a real upstream model that occupies an alias id wins
+ * and suppresses the projection.
+ */
+export declare function projectOpenAIProReasoningAliases(models: readonly ModelSpec<Api>[]): ModelSpec<Api>[];
 export interface GroqModelManagerConfig {
     apiKey?: string;
     baseUrl?: string;

package/dist/types/types.d.ts

@@ -563,6 +563,13 @@ export interface Model<TApi extends Api = Api> {
      * everything local (selection, caching, usage attribution) keys on `id`.
      */
     requestModelId?: string;
+    /**
+     * `reasoning.mode` to send on OpenAI Responses-family requests. Set on
+     * generated pro aliases (`gpt-5.6-*-pro` on `openai`/`openai-codex`) that
+     * pair a base wire id (`requestModelId`) with OpenAI's pro reasoning
+     * serving path. Absent everywhere else; providers omit the wire field.
+     */
+    reasoningMode?: "pro";
     name: string;
     api: TApi;
     provider: Provider;

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-catalog",
-	"version": "16.3.13",
+	"version": "16.3.15",
 	"description": "Model catalog for omp: bundled model database, provider discovery descriptors, model identity, classification, and equivalence",
 	"homepage": "https://omp.sh",
 	"author": "Can Boluk",
@@ -34,12 +34,12 @@
 	},
 	"dependencies": {
 		"@bufbuild/protobuf": "^2.12.0",
-		"@oh-my-pi/pi-utils": "16.3.13",
+		"@oh-my-pi/pi-utils": "16.3.15",
 		"arktype": "^2.2.0",
 		"zod": "^4"
 	},
 	"devDependencies": {
-		"@oh-my-pi/pi-ai": "16.3.13",
+		"@oh-my-pi/pi-ai": "16.3.15",
 		"@types/bun": "^1.3.14"
 	},
 	"engines": {

package/src/compat/openai.ts

@@ -541,7 +541,7 @@ export function buildOpenAICompat(spec: ModelSpec<"openai-completions">): Resolv
 			MINIMAX_PROVIDER_OR_ID_PATTERN.test(provider) || MINIMAX_PROVIDER_OR_ID_PATTERN.test(spec.id),
 		emptyLengthFinishIsContextError: provider === "ollama",
 		usesOpenAIToolCallIdLimit: provider === "openai",
-		promptCacheSessionHeader: undefined,
+		promptCacheSessionHeader: isGrok ? "x-grok-conv-id" : undefined,
 		dropThinkingWhenReasoningEffort: provider === "fireworks",
 	};
 

package/src/identity/family.ts

@@ -78,7 +78,7 @@ export const isMimoModelIdOrName = memo((value: string): boolean => {
 	return value.toLowerCase().includes("mimo");
 });
 
-const GROK_EFFORT_CAPABLE_PREFIXES = ["grok-3-mini", "grok-4.20-multi-agent", "grok-4.3"] as const;
+const GROK_EFFORT_CAPABLE_PREFIXES = ["grok-3-mini", "grok-4.20-multi-agent", "grok-4.3", "grok-4.5"] as const;
 
 /**
  * Grok SKUs that expose the wire `reasoning.effort` dial. Other Grok reasoners

package/src/model-thinking.ts

@@ -17,6 +17,7 @@ import {
 	type ParsedModel,
 	parseAnthropicModel,
 	parseKnownModel,
+	parseOpenAIModel,
 	semverEqual,
 	semverGte,
 } from "./identity/classify";
@@ -101,12 +102,14 @@ const MIMO_REASONING_EFFORT_MAP: Readonly<EffortMap> = {
 };
 
 /**
- * Effort → wire-value map for the 5-tier adaptive scale (Opus 4.7+ and
- * Fable/Mythos 5 on the Messages API). User-facing efforts shift up one notch
- * so the top tier reaches the genuine "max" and "high" lands on Anthropic's
- * recommended "xhigh" coding/agentic default.
+ * Effort → wire-value map for a shifted five-tier scale (`low..max`):
+ * user-facing efforts shift up one notch so the top tier reaches the genuine
+ * "max" and "high" lands on the recommended "xhigh" coding/agentic default.
+ * Used by Anthropic adaptive models with a real xhigh tier (Opus 4.7+ and
+ * Fable/Mythos 5 on the Messages API) and by GPT-5.6+ wire-effort models,
+ * which expose the same genuine `max` tier above `xhigh`.
  */
-export const ANTHROPIC_ADAPTIVE_EFFORT_MAP_5_TIER: Readonly<Partial<Record<Effort, string>>> = {
+export const SHIFTED_FIVE_TIER_EFFORT_MAP: Readonly<Partial<Record<Effort, string>>> = {
 	[Effort.Minimal]: "low",
 	[Effort.Low]: "medium",
 	[Effort.Medium]: "high",
@@ -295,6 +298,27 @@ function isOpenAICompatReasoningApi(api: Api): boolean {
 	return api === "openai-completions" || api === "openrouter";
 }
 
+/**
+ * GPT-5.6+ addressed through a wire `reasoning.effort`/`reasoning_effort`
+ * field, where the shifted five-tier map applies. Devin (`devin-agent`)
+ * selects effort by routing to per-tier sibling model ids instead and must
+ * stay unmapped.
+ */
+function isGpt56PlusWireEffortModel<TApi extends Api>(spec: ModelSpec<TApi>): boolean {
+	switch (spec.api) {
+		case "openai-responses":
+		case "openai-codex-responses":
+		case "azure-openai-responses":
+		case "openai-completions":
+		case "openrouter":
+			break;
+		default:
+			return false;
+	}
+	const parsed = parseOpenAIModel(bareModelId(spec.id));
+	return parsed !== null && semverGte(parsed.version, "5.6");
+}
+
 function getModelDefinedEfforts<TApi extends Api>(
 	spec: ModelSpec<TApi>,
 	compat: CompatOf<TApi>,
@@ -313,6 +337,12 @@ function getModelDefinedEfforts<TApi extends Api>(
 	if (isSakanaFuguReasoningModel(spec)) {
 		return FUGU_REASONING_EFFORTS;
 	}
+	if (isGpt56PlusWireEffortModel(spec)) {
+		// Normalize stale baked/discovered `low..xhigh` surfaces to the full
+		// five-tier ladder so the shifted map keeps the native `low` tier
+		// reachable (user `minimal`).
+		return DEFAULT_REASONING_EFFORTS_WITH_XHIGH;
+	}
 	return isOpenAICompatReasoningApi(spec.api) &&
 		(isMinimaxM2FamilyModelId(spec.id) ||
 			isOpenAIGptOssModelId(spec.id) ||
@@ -373,7 +403,7 @@ function inferDetectedEffortMap<TApi extends Api>(
 			return MINIMAX_ANTHROPIC_ADAPTIVE_EFFORT_MAP;
 		}
 		return anthropicModelHasRealXHighEffort(spec, parsedModel)
-			? ANTHROPIC_ADAPTIVE_EFFORT_MAP_5_TIER
+			? SHIFTED_FIVE_TIER_EFFORT_MAP
 			: ANTHROPIC_ADAPTIVE_EFFORT_MAP_4_TIER;
 	}
 	// GLM-5.2 coding SKUs accept `reasoning_effort`, but the effort dialect is
@@ -397,6 +427,9 @@ function inferDetectedEffortMap<TApi extends Api>(
 	if (isSakanaFuguReasoningModel(spec)) {
 		return FUGU_REASONING_EFFORT_MAP;
 	}
+	if (isGpt56PlusWireEffortModel(spec)) {
+		return SHIFTED_FIVE_TIER_EFFORT_MAP;
+	}
 	if (!isOpenAICompatReasoningApi(spec.api)) {
 		return undefined;
 	}
@@ -446,7 +479,7 @@ function getOpenRouterAnthropicReasoningEffortMap(modelId: string): EffortMap |
 	if (!isAnthropicAdaptiveGenAtLeast(parsed, "4.6")) return undefined;
 
 	const hasRealXHigh = isAnthropicAdaptiveGenAtLeast(parsed, "4.7");
-	return hasRealXHigh ? ANTHROPIC_ADAPTIVE_EFFORT_MAP_5_TIER : ANTHROPIC_ADAPTIVE_EFFORT_MAP_4_TIER;
+	return hasRealXHigh ? SHIFTED_FIVE_TIER_EFFORT_MAP : ANTHROPIC_ADAPTIVE_EFFORT_MAP_4_TIER;
 }
 
 function inferSupportedEfforts<TApi extends Api>(
@@ -474,6 +507,11 @@ function inferOpenAISupportedEfforts(model: OpenAIModel): readonly Effort[] {
 	if (model.variant === "codex-mini" && semverEqual(model.version, "5.1")) {
 		return GPT_5_1_CODEX_MINI_EFFORTS;
 	}
+	// 5.6+ exposes the full five-tier ladder: the shifted wire map spans
+	// low..max, with user `minimal` reaching the native `low` tier.
+	if (semverGte(model.version, "5.6")) {
+		return DEFAULT_REASONING_EFFORTS_WITH_XHIGH;
+	}
 	if (semverGte(model.version, "5.2")) {
 		return GPT_5_2_PLUS_EFFORTS;
 	}