feat(glm): add GLM-5.2 with effort-tier routing (high/max)

config/quality/file-size-baseline.json

@@ -2,6 +2,7 @@
   "_comment": "Catraca de tamanho (check-file-size.mjs). frozen so pode encolher; arquivos novos <= cap. --update ratcheta.",
   "_rebaseline_v3.8.25": "Drift consciente do ciclo v3.8.24->v3.8.25 (features #3799-#3806: free-provider-rankings, plugins menu, proxy IP-family selector). 3 arquivos cresceram por feature legitima, nao por regressao de qualidade: ProxyRegistryManager.tsx 1072->1089, sidebarVisibility.ts 990->1006, schemas.ts 2519->2522. Encolher fica como debt para um refactor dedicado.",
   "_rebaseline_2026_06_15_3860_compression_ui": "PR #3860 own growth: sidebarVisibility.ts 1006->1100 (+94 = Compression Hub menu entries: Hub + per-engine Lite/Aggressive/Ultra pages + combos editor) and chatCore.ts 5812->5815 (+3 = compression UI config wiring). Cohesive feature growth, not a quality regression.",
+  "_rebaseline_2026_06_15_3885_glm_5_2": "PR #3885 own growth: pricing.ts 1508->1529 (+21 = GLM-5.2 pricing rows for glm-5.2 + effort aliases glm-5.2-high/-max, same $1.2/$5 schedule as glm-5.1; pure data). Also adds glm-5.2 specs to glmProvider.ts/modelSpecs.ts (modelSpecs.ts stays under cap). Cohesive model registration; not extractable.",
   "_rebaseline_2026_06_15_3870_alias_lookup": "PR #3870 own growth: providerRegistry.ts 4703->4708 (+5 = generateModels() also stores each provider's models under its raw id, not only its alias, so getProviderModels(rawId) works when alias != id e.g. github->gh; preserves the existing first-wins guard). Cohesive registry fix; not extractable.",
   "_rebaseline_2026_06_15_3871_empty_pool": "PR #3871 own growth: combo.ts 5203->5204 (+1 = guard expandAutoComboCandidatePool against an empty candidatePool array — Array.isArray(pool) && pool.length > 0 so [] falls through to active-connection expansion instead of early-returning). One-line correctness fix; not extractable.",
   "cap": 800,
@@ -101,7 +102,7 @@
     "src/shared/components/RequestLoggerV2.tsx": 1282,
     "src/shared/components/analytics/charts.tsx": 1558,
     "src/shared/constants/cliTools.ts": 875,
-    "src/shared/constants/pricing.ts": 1508,
+    "src/shared/constants/pricing.ts": 1529,
     "src/shared/constants/providers.ts": 3147,
     "src/shared/constants/sidebarVisibility.ts": 1100,
     "src/shared/services/cliRuntime.ts": 1090,

open-sse/config/glmProvider.ts

@@ -16,6 +16,30 @@ export const GLM_ANTHROPIC_DEFAULT_BASE_URLS = Object.freeze({
 });
 
 export const GLM_SHARED_MODELS = Object.freeze([
+  {
+    id: "glm-5.2",
+    name: "GLM 5.2",
+    contextLength: 1000000,
+    maxOutputTokens: 131072,
+    toolCalling: true,
+    supportsReasoning: true,
+  },
+  {
+    id: "glm-5.2-high",
+    name: "GLM 5.2 High",
+    contextLength: 1000000,
+    maxOutputTokens: 131072,
+    toolCalling: true,
+    supportsReasoning: true,
+  },
+  {
+    id: "glm-5.2-max",
+    name: "GLM 5.2 Max",
+    contextLength: 1000000,
+    maxOutputTokens: 131072,
+    toolCalling: true,
+    supportsReasoning: true,
+  },
   {
     id: "glm-5.1",
     name: "GLM 5.1",

open-sse/executors/glm.ts

@@ -50,6 +50,19 @@ function getEffectiveKey(credentials: ProviderCredentials): string {
   return credentials.apiKey || credentials.accessToken || "";
 }
 
+/**
+ * GLM-5.2 effort tiers route exclusively through the Anthropic transport,
+ * where Zhipu maps Claude Code effort selectors (high/max) to reasoning
+ * intensity. The base model ID sent upstream is always "glm-5.2".
+ *
+ * https://docs.z.ai/devpack/latest-model
+ */
+function parseGlm52Effort(model: string): { baseModel: string; effort: "high" | "max" } | null {
+  if (model === "glm-5.2-high") return { baseModel: "glm-5.2", effort: "high" };
+  if (model === "glm-5.2-max") return { baseModel: "glm-5.2", effort: "max" };
+  return null;
+}
+
 function applyGlmRequestDefaults(body: unknown, defaults?: JsonRecord | null): unknown {
   const record = asRecord(body);
   if (!record || !defaults) return body;
@@ -228,27 +241,61 @@ export class GlmExecutor extends DefaultExecutor {
     credentials: ProviderCredentials,
     transport: GlmTransport
   ) {
-    const transformed = this.transformRequest(model, body, stream, credentials);
+    const effortTier = parseGlm52Effort(model);
+    const effectiveModel = effortTier ? effortTier.baseModel : model;
+
+    const transformed = this.transformRequest(effectiveModel, body, stream, credentials);
+    const record = asRecord(transformed);
+
+    // Ensure upstream receives the base model ID, not the effort-suffixed alias
+    if (record && effortTier) {
+      record.model = effectiveModel;
+    }
 
     if (transport === "openai") {
-      const record = asRecord(transformed);
       if (record && stream && hasTools(record) && record.tool_stream === undefined) {
         return { ...record, tool_stream: true };
       }
       return transformed;
     }
 
-    return translateRequest(
+    const translated = translateRequest(
       FORMATS.OPENAI,
       FORMATS.CLAUDE,
-      model,
-      { ...(transformed as JsonRecord), _disableToolPrefix: true },
+      effectiveModel,
+      { ...(record ?? {}), _disableToolPrefix: true },
       stream,
       credentials,
       this.provider,
       null,
       { preserveCacheControl: false }
     );
+
+    // Inject effort and thinking for the Anthropic transport.
+    // Zhipu's Anthropic endpoint requires thinking.type=enabled to emit
+    // thinking_delta blocks in the SSE response. Without it, reasoning is
+    // not surfaced and clients see no thinking content.
+    // The effort-2025-11-24 beta header (in GLM_ANTHROPIC_BETA) carries
+    // the high/max intensity selector.
+    if (effortTier) {
+      const translatedRecord = asRecord(translated);
+      if (translatedRecord) {
+        translatedRecord.effort = effortTier.effort;
+        // Zhipu's Anthropic endpoint only supports thinking.type
+        // "enabled"/"disabled" — not "adaptive". Clients like Claude Code
+        // default to "adaptive" for reasoning models, so force "enabled"
+        // here while preserving any other fields (e.g. budget_tokens).
+        const existingThinking = asRecord(translatedRecord.thinking);
+        if (!existingThinking || existingThinking.type !== "enabled") {
+          translatedRecord.thinking = {
+            ...existingThinking,
+            type: "enabled",
+          };
+        }
+      }
+    }
+
+    return translated;
   }
 
   private async executeTransport(
@@ -343,6 +390,15 @@ export class GlmExecutor extends DefaultExecutor {
   }
 
   async execute(input: ExecuteInput): Promise<GlmExecuteResult> {
+    const effortTier = parseGlm52Effort(input.model);
+
+    // GLM-5.2 effort tiers route directly through Anthropic transport (no fallback).
+    // Zhipu only graduates effort on the Anthropic endpoint via the
+    // effort-2025-11-24 beta header included in GLM_ANTHROPIC_BETA.
+    if (effortTier) {
+      return this.executeTransport(input, "anthropic");
+    }
+
     const primaryTransport = getGlmTransport(
       input.credentials.providerSpecificData,
       this.config.baseUrl

open-sse/mcp-server/__tests__/glmCodingProviderConfig.test.ts

@@ -88,6 +88,9 @@ describe("GLM Coding provider registry surfaces", () => {
     expect(PROVIDER_ID_TO_ALIAS.glm).toBe("glm");
     expect(byProviderId).toEqual(byAlias);
     expect(byProviderId.map((model) => model.id)).toEqual([
+      "glm-5.2",
+      "glm-5.2-high",
+      "glm-5.2-max",
       "glm-5.1",
       "glm-5",
       "glm-5-turbo",
@@ -101,6 +104,30 @@ describe("GLM Coding provider registry surfaces", () => {
     ]);
   });
 
+  it("registers GLM-5.2 with correct specs and effort tier aliases", () => {
+    const models = getModelsByProviderId("glm");
+    const get = (id: string) => models.find((m) => m.id === id);
+
+    // Base model
+    const base = get("glm-5.2");
+    expect(base).toBeDefined();
+    expect(base?.contextLength).toBe(1000000);
+    expect(base?.maxOutputTokens).toBe(131072);
+    expect(base?.supportsReasoning).toBe(true);
+    expect(base?.toolCalling).toBe(true);
+
+    // Effort tier aliases share the same specs
+    const high = get("glm-5.2-high");
+    expect(high).toBeDefined();
+    expect(high?.contextLength).toBe(1000000);
+    expect(high?.maxOutputTokens).toBe(131072);
+
+    const max = get("glm-5.2-max");
+    expect(max).toBeDefined();
+    expect(max?.contextLength).toBe(1000000);
+    expect(max?.maxOutputTokens).toBe(131072);
+  });
+
   it("applies doc-backed context window overrides for GLM models", () => {
     const models = getModelsByProviderId("glm");
     const get = (id: string) => models.find((m) => m.id === id);
@@ -126,6 +153,9 @@ describe("GLM Coding provider registry surfaces", () => {
     expect(supportsToolCalling("glm/glm-5")).toBe(true);
     expect(supportsToolCalling("glm/glm-4.7-flash")).toBe(true);
     expect(supportsToolCalling("glm/glm-4.5-air")).toBe(true);
+    expect(supportsToolCalling("glm/glm-5.2")).toBe(true);
+    expect(supportsToolCalling("glm/glm-5.2-high")).toBe(true);
+    expect(supportsToolCalling("glm/glm-5.2-max")).toBe(true);
 
     expect(getPricingForModel("glm", "glm-5")).toEqual({
       input: 1.0,
@@ -148,6 +178,20 @@ describe("GLM Coding provider registry surfaces", () => {
       reasoning: 1.1,
       cache_creation: 0.2,
     });
+    expect(getPricingForModel("glm", "glm-5.2")).toEqual({
+      input: 1.2,
+      output: 5,
+      cached: 0.3,
+      reasoning: 5,
+      cache_creation: 1.2,
+    });
+    expect(getPricingForModel("glm", "glm-5.2-max")).toEqual({
+      input: 1.2,
+      output: 5,
+      cached: 0.3,
+      reasoning: 5,
+      cache_creation: 1.2,
+    });
   });
 
   it("keeps the repo-derived GLM inventory internally aligned across registry and pricing surfaces", () => {

src/shared/constants/modelSpecs.ts

@@ -354,6 +354,26 @@ export const MODEL_SPECS: Record<string, ModelSpec> = {
     supportsTools: true,
   },
 
+  // ── Z.AI GLM-5.2 (1M context, 128K max output, effort tiers) ────
+  "glm-5.2": {
+    maxOutputTokens: 131072,
+    contextWindow: 1000000,
+    supportsThinking: true,
+    supportsTools: true,
+  },
+  "glm-5.2-high": {
+    maxOutputTokens: 131072,
+    contextWindow: 1000000,
+    supportsThinking: true,
+    supportsTools: true,
+  },
+  "glm-5.2-max": {
+    maxOutputTokens: 131072,
+    contextWindow: 1000000,
+    supportsThinking: true,
+    supportsTools: true,
+  },
+
   // ── Z.AI GLM-5.x (200K context, 128K max output) ─────────────────
   "glm-5.1": {
     maxOutputTokens: 128000,

src/shared/constants/pricing.ts

@@ -60,6 +60,27 @@ const CLAUDE_SONNET_46_PRICING = {
 };
 
 const GLM_PRICING = {
+  "glm-5.2": {
+    input: 1.2,
+    output: 5,
+    cached: 0.3,
+    reasoning: 5,
+    cache_creation: 1.2,
+  },
+  "glm-5.2-high": {
+    input: 1.2,
+    output: 5,
+    cached: 0.3,
+    reasoning: 5,
+    cache_creation: 1.2,
+  },
+  "glm-5.2-max": {
+    input: 1.2,
+    output: 5,
+    cached: 0.3,
+    reasoning: 5,
+    cache_creation: 1.2,
+  },
   "glm-5.1": {
     input: 1.2,
     output: 5,