🤖 feat: add Z.ai as a native provider with GLM 5.3 Flash

docs/config/models.mdx

@@ -34,6 +34,7 @@ Xum ships with curated models kept up to date with the frontier. Use any custom
 | DeepSeek V4 Pro        | deepseek:deepseek-v4-pro      | `deepseek`, `deepseek-pro`, `deepseek-v4`, `deepseek-v4-pro` |         |
 | DeepSeek V4 Flash      | deepseek:deepseek-v4-flash    | `deepseek-flash`, `deepseek-v4-flash`                        |         |
 | Kimi K3                | moonshotai:kimi-k3            | `kimi`, `k3`, `kimi-k3`                                      |         |
+| GLM 5.3 Flash          | zai:glm-5.3-flash             | `glm`, `glm-flash`, `glm-5.3-flash`                          |         |
 
 {/* END KNOWN_MODELS_TABLE */}
 

docs/config/providers.mdx

@@ -24,6 +24,7 @@ Most providers only need an API key. The UI handles validation and shows which p
 | **xAI**            | Grok                        | [console.x.ai](https://console.x.ai/)                   |
 | **DeepSeek**       | DeepSeek Chat, Reasoner     | [platform.deepseek.com](https://platform.deepseek.com/) |
 | **Moonshot AI**    | Kimi K3                     | [platform.moonshot.ai](https://platform.moonshot.ai/)   |
+| **Z.ai**           | GLM 5.3 Flash               | [z.ai](https://z.ai/model-api)                          |
 | **OpenRouter**     | 300+ models                 | [openrouter.ai](https://openrouter.ai/)                 |
 | **Ollama**         | Local models                | [ollama.com](https://ollama.com/) (no key needed)       |
 | **Bedrock**        | Claude via AWS              | AWS Console                                             |
@@ -45,6 +46,7 @@ Providers also read from environment variables as fallback:
 | OpenRouter     | `OPENROUTER_API_KEY`                               |
 | DeepSeek       | `DEEPSEEK_API_KEY`                                 |
 | Moonshot AI    | `MOONSHOT_API_KEY`                                 |
+| Z.ai           | `ZAI_API_KEY`                                      |
 | github-copilot | `GITHUB_COPILOT_TOKEN`                             |
 | Bedrock        | `AWS_REGION` (credentials via AWS SDK chain)       |
 

flake.nix

@@ -84,7 +84,7 @@
 
             outputHashMode = "recursive";
             # Marker used by scripts/update_flake_hash.sh to update this hash in place.
-            outputHash = "sha256-ri3Q1gY4ifnjT9FMg3wNqoIT5OJbfQRKAWw5Zt8DH9k="; # xum-offline-cache-hash
+            outputHash = "sha256-qZmS1EPW+0IU6S8cQGuj+wjHTgBRB0HhLkM0cJDgoc8="; # xum-offline-cache-hash
           };
 
           configurePhase = ''

package.json

@@ -57,6 +57,7 @@
     "@ai-sdk/openai": "^4.0.43",
     "@ai-sdk/openai-compatible": "^3.0.7",
     "@ai-sdk/xai": "^4.0.37",
+    "@ai-sdk/zai": "^3.0.2",
     "@aws-sdk/credential-providers": "^3.940.0",
     "@dnd-kit/core": "^6.3.1",
     "@dnd-kit/sortable": "^10.0.0",

scripts/gen_docs.ts

@@ -370,6 +370,7 @@ const PROVIDER_DISPLAY_NAMES: Record<string, string> = {
   xai: "xAI",
   deepseek: "DeepSeek",
   moonshotai: "Moonshot AI",
+  zai: "Z.ai",
   openrouter: "OpenRouter",
   bedrock: "Bedrock",
 };

src/browser/components/ProjectSidebar/ProjectSidebar.test.tsx

@@ -63,6 +63,7 @@ function installProviderIconSvgMocks() {
     "@/browser/assets/icons/ollama.svg?react",
     "@/browser/assets/icons/deepseek.svg?react",
     "@/browser/assets/icons/moonshotai.svg?react",
+    "@/browser/assets/icons/zai.svg?react",
     "@/browser/assets/icons/aws.svg?react",
     "@/browser/assets/icons/github.svg?react",
     "@/browser/assets/icons/coder.svg?react",

src/browser/components/ProviderIcon/ProviderIcon.tsx

@@ -8,6 +8,7 @@ import OpenRouterIcon from "@/browser/assets/icons/openrouter.svg?react";
 import OllamaIcon from "@/browser/assets/icons/ollama.svg?react";
 import DeepSeekIcon from "@/browser/assets/icons/deepseek.svg?react";
 import MoonshotAIIcon from "@/browser/assets/icons/moonshotai.svg?react";
+import ZaiIcon from "@/browser/assets/icons/zai.svg?react";
 import AWSIcon from "@/browser/assets/icons/aws.svg?react";
 import GitHubIcon from "@/browser/assets/icons/github.svg?react";
 import CoderIcon from "@/browser/assets/icons/coder.svg?react";
@@ -30,6 +31,7 @@ const PROVIDER_ICONS: Partial<Record<ProviderName, React.FC>> = {
   xai: XAIIcon,
   deepseek: DeepSeekIcon,
   moonshotai: MoonshotAIIcon,
+  zai: ZaiIcon,
   openrouter: OpenRouterIcon,
   bedrock: AWSIcon,
   ollama: OllamaIcon,

src/browser/features/Settings/Sections/ProvidersSection.tsx

@@ -265,6 +265,7 @@ const PROVIDER_KEY_URLS: Partial<Record<ProviderName, string>> = {
   xai: "https://console.x.ai/team/default/api-keys",
   deepseek: "https://platform.deepseek.com/api_keys",
   moonshotai: "https://platform.moonshot.ai/console/api-keys",
+  zai: "https://z.ai/manage-apikey/apikey-list",
   openrouter: "https://openrouter.ai/settings/keys",
   // bedrock: AWS credential chain, no simple key URL
   // ollama: local service, no key needed

src/common/config/schemas/providersConfig.ts

@@ -139,6 +139,7 @@ export const CoderProviderConfigSchema = BaseProviderConfigSchema.extend({
 export const GoogleProviderConfigSchema = BaseProviderConfigSchema;
 export const DeepSeekProviderConfigSchema = BaseProviderConfigSchema;
 export const MoonshotAIProviderConfigSchema = BaseProviderConfigSchema;
+export const ZaiProviderConfigSchema = BaseProviderConfigSchema;
 export const OllamaProviderConfigSchema = BaseProviderConfigSchema;
 export const GitHubCopilotProviderConfigSchema = BaseProviderConfigSchema;
 
@@ -153,6 +154,7 @@ export const ProvidersConfigSchema = z
     google: GoogleProviderConfigSchema.optional(),
     deepseek: DeepSeekProviderConfigSchema.optional(),
     moonshotai: MoonshotAIProviderConfigSchema.optional(),
+    zai: ZaiProviderConfigSchema.optional(),
     ollama: OllamaProviderConfigSchema.optional(),
     "github-copilot": GitHubCopilotProviderConfigSchema.optional(),
     coder: CoderProviderConfigSchema.optional(),

src/common/constants/knownModels.test.ts

@@ -49,6 +49,14 @@ describe("Known Models Integration", () => {
     expect(MODEL_ABBREVIATIONS.k3).toBe("moonshotai:kimi-k3");
   });
 
+  test("glm aliases resolve only to the direct Z.ai GLM 5.3 Flash model", () => {
+    expect(MODEL_ABBREVIATIONS.glm).toBe("zai:glm-5.3-flash");
+    expect(MODEL_ABBREVIATIONS["glm-flash"]).toBe("zai:glm-5.3-flash");
+    expect(Object.values(KNOWN_MODELS).filter((model) => model.provider === "zai")).toEqual([
+      KNOWN_MODELS.GLM_53_FLASH,
+    ]);
+  });
+
   test("known model ids and aliases stay unique across the curated registry", () => {
     const seenIds = new Set<string>();
     const seenAliases = new Set<string>();

src/common/constants/knownModels.ts

@@ -4,7 +4,7 @@
 
 import { formatModelDisplayName } from "../utils/ai/modelDisplay";
 
-type ModelProvider = "anthropic" | "openai" | "google" | "xai" | "deepseek" | "moonshotai";
+type ModelProvider = "anthropic" | "openai" | "google" | "xai" | "deepseek" | "moonshotai" | "zai";
 
 interface KnownModelDefinition {
   /** Provider identifier used by SDK factories */
@@ -225,6 +225,13 @@ const MODEL_DEFINITIONS = {
     // Moonshot tokenizer available) for approximate counting.
     tokenizerOverride: "moonshotai/kimi-k2",
   },
+  GLM_53_FLASH: {
+    provider: "zai",
+    providerModelId: "glm-5.3-flash",
+    aliases: ["glm", "glm-flash", "glm-5.3-flash"],
+    // GLM 5.3 is not in ai-tokenizer yet; use the closest published GLM encoding.
+    tokenizerOverride: "zai/glm-4.5",
+  },
 } as const satisfies Record<string, KnownModelDefinition>;
 
 export type KnownModelKey = keyof typeof MODEL_DEFINITIONS;

src/common/constants/providers.ts

@@ -21,6 +21,7 @@ export type ProviderName =
   | "xai"
   | "deepseek"
   | "moonshotai"
+  | "zai"
   | "openrouter"
   | "github-copilot"
   | "coder"
@@ -158,12 +159,20 @@ export const PROVIDER_DEFINITIONS = {
     requiresApiKey: true,
     kind: "direct",
   },
+  zai: {
+    displayName: "Z.ai",
+    import: () => import("@ai-sdk/zai"),
+    factoryName: "createZai",
+    requiresApiKey: true,
+    kind: "direct",
+  },
   openrouter: {
     displayName: "OpenRouter",
     import: () => import("@openrouter/ai-sdk-provider"),
     factoryName: "createOpenRouter",
     requiresApiKey: true,
     kind: "gateway",
+    // OpenRouter uses the hyphenated vendor slug "z-ai", so direct "zai" routing is not inferred.
     routes: ["anthropic", "openai", "google", "xai", "deepseek", "moonshotai"],
     passthrough: false,
     toGatewayModelId: toSlashSeparatedGatewayModelId,

src/common/types/thinking.ts

@@ -325,6 +325,12 @@ export function isGrok46Model(modelString: string): boolean {
   return /^grok-4\.6(?:$|-)/.test(withoutPrefix);
 }
 
+/** GLM 5.3 and GLM 5.3 Flash always reason with low, high, or max effort. */
+export function isGlm53Model(modelString: string): boolean {
+  const withoutPrefix = stripModelProviderPrefixes(modelString);
+  return /^glm-5\.3(?:-flash)?(?:$|-(?:latest|\d))/.test(withoutPrefix);
+}
+
 /**
  * Kimi K3 (Moonshot AI) always reasons and supports only the max reasoning
  * effort; the thinking policy and the Moonshot/OpenRouter provider-options

src/common/utils/ai/modelCapabilities.test.ts

@@ -46,6 +46,15 @@ describe("getModelCapabilities", () => {
     }
   });
 
+  it("reports image-only multimodal input for GLM 5.3 Flash", () => {
+    const caps = getModelCapabilities("zai:glm-5.3-flash");
+
+    expect(caps).not.toBeNull();
+    expect(caps?.supportsVision).toBe(true);
+    expect(caps?.supportsPdfInput).toBe(false);
+    expect(caps?.supportsVideoInput).toBe(false);
+  });
+
   it("infers PDF support for OpenAI vision models when models-extra omits the flag", () => {
     const caps = getModelCapabilities("openai:gpt-5.5");
     expect(caps).not.toBeNull();

src/common/utils/ai/modelDisplay.test.ts

@@ -31,6 +31,12 @@ describe("formatModelDisplayName", () => {
     });
   });
 
+  describe("GLM models", () => {
+    test("preserves uppercase GLM branding", () => {
+      expect(formatModelDisplayName("glm-5.3-flash")).toBe("GLM 5.3 Flash");
+    });
+  });
+
   describe("GPT models", () => {
     test("formats GPT models", () => {
       expect(formatModelDisplayName("gpt-5-pro")).toBe("GPT-5 Pro");

src/common/utils/ai/modelDisplay.ts

@@ -81,6 +81,12 @@ export function formatModelDisplayName(modelName: string): string {
     }
   }
 
+  // GLM models
+  if (lower.startsWith("glm-")) {
+    const parts = lower.slice("glm-".length).split("-");
+    return `GLM ${parts.map(capitalize).join(" ")}`;
+  }
+
   // GPT models
   if (lower.startsWith("gpt-")) {
     const parts = lower.split("-");

src/common/utils/ai/providerOptions.test.ts

@@ -1862,6 +1862,28 @@ describe("buildProviderOptions - Moonshot", () => {
   });
 });
 
+describe("buildProviderOptions - Z.ai", () => {
+  test("enables thinking and maps every GLM 5.3 policy level", () => {
+    for (const [level, reasoningEffort] of [
+      ["low", "low"],
+      ["high", "high"],
+      ["max", "max"],
+    ] as const) {
+      expect(buildProviderOptions("zai:glm-5.3-flash", level)).toEqual({
+        zai: {
+          thinking: { type: "enabled" },
+          reasoningEffort,
+          toolStream: true,
+        },
+      });
+    }
+  });
+
+  test("does not apply forced GLM 5.3 options to other Z.ai models", () => {
+    expect(buildProviderOptions("zai:glm-4.7-flash", "high")).toEqual({});
+  });
+});
+
 describe("buildProviderOptions - OpenRouter", () => {
   test("sends the explicit max effort for OpenRouter-routed Kimi K3", () => {
     // `enabled: true` alone falls back to OpenRouter's default (medium) effort,

src/common/utils/ai/providerOptions.ts

@@ -10,6 +10,7 @@ import type { AnthropicProviderOptions } from "@ai-sdk/anthropic";
 import type { GoogleGenerativeAIProviderOptions } from "@ai-sdk/google";
 import type { OpenAIResponsesProviderOptions } from "@ai-sdk/openai";
 import type { JSONValue } from "@ai-sdk/provider";
+import type { ZaiLanguageModelChatOptions } from "@ai-sdk/zai";
 import type {
   XaiProviderOptions,
   // Chat options alias does not include store; Responses options do (frontier Grok / ZDR).
@@ -28,6 +29,7 @@ import {
   getOpenAIReasoningEffort,
   isGrok46Model,
   isGrokFrontierModel,
+  isGlm53Model,
   isKimiK3Model,
   openaiSupportsProMode,
   OPENROUTER_REASONING_EFFORT,
@@ -147,6 +149,7 @@ type ProviderOptions =
   | { google: GoogleGenerativeAIProviderOptions }
   | { openrouter: OpenRouterReasoningOptions }
   | { moonshotai: MoonshotAIProviderOptions }
+  | { zai: ZaiLanguageModelChatOptions }
   | { xai: XaiBuiltProviderOptions }
   | { "github-copilot": OpenAICompatibleGatewayProviderOptions }
   | Record<string, never>; // Empty object for unsupported providers
@@ -590,6 +593,29 @@ export function buildProviderOptions(
     return {};
   }
 
+  // Build Z.ai-specific options
+  if (formatProvider === "zai") {
+    if (!isGlm53Model(capabilityModel)) {
+      return {};
+    }
+
+    const reasoningEffort: ZaiLanguageModelChatOptions["reasoningEffort"] =
+      effectiveThinking === "max"
+        ? "max"
+        : effectiveThinking === "high" || effectiveThinking === "xhigh"
+          ? "high"
+          : "low";
+    const options = {
+      zai: {
+        thinking: { type: "enabled" },
+        reasoningEffort,
+        toolStream: true,
+      },
+    } satisfies { zai: ZaiLanguageModelChatOptions };
+    log.debug("buildProviderOptions: Returning Z.ai options", options);
+    return options;
+  }
+
   // Build OpenRouter-specific options
   if (formatProvider === "openrouter") {
     // Kimi K3 always reasons and supports only the max reasoning effort. Send it

src/common/utils/thinking/policy.test.ts

@@ -458,7 +458,7 @@ describe("getThinkingPolicyForModel", () => {
     expect(enforceThinkingPolicy("anthropic:claude-mythos-5", "off")).toBe("low");
   });
 
-  test("resolveEffectiveThinkingLevel clamps unset/off for Mythos-class only", () => {
+  test("resolveEffectiveThinkingLevel clamps unset/off for forced-thinking models", () => {
     // Mythos-class cannot disable thinking: unset and "off" both resolve to "low"
     // so provider options, replay transforms, and metadata stay consistent with
     // the provider's always-thinking behavior.
@@ -956,6 +956,29 @@ describe("Grok 4.6 thinking policy", () => {
   });
 });
 
+describe("GLM 5.3 thinking policy", () => {
+  test("offers only Z.ai's forced-thinking effort levels", () => {
+    for (const model of ["glm-5.3-flash", "zai:glm-5.3", "zai:glm-5.3-flash-2026-08-26"]) {
+      expect(getThinkingPolicyForModel(model)).toEqual(["low", "high", "max"]);
+    }
+
+    expect(enforceThinkingPolicy("zai:glm-5.3-flash", "off")).toBe("low");
+    expect(enforceThinkingPolicy("zai:glm-5.3-flash", "medium")).toBe("low");
+    expect(enforceThinkingPolicy("zai:glm-5.3-flash", "xhigh")).toBe("high");
+    expect(resolveEffectiveThinkingLevel("zai:glm-5.3-flash", undefined)).toBe("low");
+    expect(resolveEffectiveThinkingLevel("zai:glm-5.3-flash", "off")).toBe("low");
+  });
+
+  test("does not apply the GLM 5.3 policy to named variants", () => {
+    expect(getThinkingPolicyForModel("zai:glm-5.3-flashx")).toEqual([
+      "off",
+      "low",
+      "medium",
+      "high",
+    ]);
+  });
+});
+
 describe("getAvailableThinkingLevels", () => {
   test("returns the raw capability when no floor is provided", () => {
     expect(getAvailableThinkingLevels("anthropic:claude-sonnet-4-5")).toEqual([