feat: model updates and + z.ai (#10638)

* Add latest model definitions across providers

Add new models for OpenAI (gpt-4.1, gpt-5.2, codex-mini, o1/o3/o4-mini),
Gemini (2.5 stable, 2.5-flash-lite, 3-flash-preview), Mistral (devstral,
magistral, ministral, mistral-nemo), xAI (grok-4-fast, grok-4-1-fast,
grok-3-fast), and Bedrock (Claude 4.5/4.6, Nova, Llama 4, DeepSeek).

Update tool support and reasoning detection for the new model families.

* Add Z.ai (GLM) provider, Kimi K2/K2.5 models, and fix OpenRouter tool support bug

- Add Z.ai as a new provider with GLM-5, GLM-4.7, GLM-4-plus, GLM-4.5 models
- Add Kimi K2 and K2.5 model packages to Moonshot provider
- Fix variable shadowing bug in OpenRouter tool support where loop
  variables shadowed the outer model parameter, causing specificModels
  and supportedContains checks to always return true
- Add moonshotai/kimi and zai-org/glm to OpenRouter supported prefixes
- Add Z.ai and Moonshot to docs navigation

* Clean up unnecessary import aliases for zAI

* Add Z.ai Coding Plan endpoint note to docs

* Replace placeholder zai.png icon with proper zAI.svg

* Add zAI provider to openai-adapters

* Fix Z.ai API key URL

Author: RomneyDa

core/llm/autodetect.ts

@@ -71,6 +71,7 @@ const PROVIDER_HANDLES_TEMPLATING: string[] = [
   "gemini",
   "docker",
   "nous",
+  "zAI",
   // TODO add these, change to inverted logic so only the ones that need templating are hardcoded
   // Asksage.ts
   // Azure.ts
@@ -128,6 +129,7 @@ const PROVIDER_SUPPORTS_IMAGES: string[] = [
   "nebius",
   "ovhcloud",
   "watsonx",
+  "zAI",
 ];
 
 const MODEL_SUPPORTS_IMAGES: RegExp[] = [
@@ -209,6 +211,19 @@ function modelSupportsReasoning(
   if (model.model.includes("deepseek-r")) {
     return true;
   }
+  // o-series reasoning models
+  if (/^o[134]/.test(model.model)) {
+    return true;
+  }
+  if (model.model.includes("codex")) {
+    return true;
+  }
+  if (model.model.includes("magistral")) {
+    return true;
+  }
+  if (model.model.includes("grok-4")) {
+    return true;
+  }
 
   return false;
 }
@@ -257,6 +272,7 @@ function isProviderHandlesTemplatingOrNoTemplateTypeRequired(
     modelName.includes("moonshot") ||
     modelName.includes("kimi") ||
     modelName.includes("mercury") ||
+    modelName.includes("glm") ||
     /^o\d/.test(modelName)
   );
 }

core/llm/llms/index.ts

@@ -67,6 +67,7 @@ import Vllm from "./Vllm";
 import Voyage from "./Voyage";
 import WatsonX from "./WatsonX";
 import xAI from "./xAI";
+import zAI from "./zAI";
 export const LLMClasses = [
   Anthropic,
   Cohere,
@@ -126,6 +127,7 @@ export const LLMClasses = [
   Voyage,
   LlamaStack,
   TARS,
+  zAI,
 ];
 
 export async function llmFromDescription(

core/llm/llms/zAI.ts

@@ -0,0 +1,13 @@
+import { LLMOptions } from "../../index.js";
+
+import OpenAI from "./OpenAI.js";
+
+class zAI extends OpenAI {
+  static providerName = "zAI";
+  static defaultOptions: Partial<LLMOptions> = {
+    apiBase: "https://api.z.ai/api/paas/v4/",
+    useLegacyCompletionsEndpoint: false,
+  };
+}
+
+export default zAI;

core/llm/toolSupport.ts

@@ -41,7 +41,11 @@ export const PROVIDER_TOOL_SUPPORT: Record<string, (model: string) => boolean> =
       if (
         lower.startsWith("gpt-4") ||
         lower.startsWith("gpt-5") ||
-        lower.startsWith("o3")
+        lower.startsWith("o1") ||
+        lower.startsWith("o3") ||
+        lower.startsWith("o4") ||
+        lower.startsWith("codex") ||
+        lower.startsWith("gpt-4.1")
       ) {
         return true;
       }
@@ -114,6 +118,9 @@ export const PROVIDER_TOOL_SUPPORT: Record<string, (model: string) => boolean> =
           "nova-micro",
           "nova-premier",
           "gpt-oss",
+          "llama4",
+          "llama-4",
+          "deepseek",
         ].some((part) => model.toLowerCase().includes(part))
       ) {
         return true;
@@ -133,7 +140,7 @@ export const PROVIDER_TOOL_SUPPORT: Record<string, (model: string) => boolean> =
           "pixtral",
           "ministral",
           "mistral-nemo",
-          "devstral",
+          "magistral",
         ].some((part) => model.toLowerCase().includes(part))
       );
     },
@@ -287,9 +294,11 @@ export const PROVIDER_TOOL_SUPPORT: Record<string, (model: string) => boolean> =
       const supportedPrefixes = [
         "openai/gpt-3.5",
         "openai/gpt-4",
+        "openai/gpt-5",
         "openai/o1",
         "openai/o3",
         "openai/o4",
+        "openai/codex",
         "openai/gpt-oss",
         "anthropic/claude",
         "microsoft/phi-3",
@@ -316,6 +325,8 @@ export const PROVIDER_TOOL_SUPPORT: Record<string, (model: string) => boolean> =
         "meta-llama/llama-4",
         "all-hands/openhands-lm-32b",
         "lgai-exaone/exaone",
+        "moonshotai/kimi",
+        "zai-org/glm",
       ];
       for (const prefix of supportedPrefixes) {
         if (model.toLowerCase().startsWith(prefix)) {
@@ -333,21 +344,25 @@ export const PROVIDER_TOOL_SUPPORT: Record<string, (model: string) => boolean> =
         "nousresearch/hermes-3-llama-3.1-70b",
         "moonshotai/kimi-k2",
       ];
-      for (const model of specificModels) {
-        if (model.toLowerCase() === model) {
+      for (const specificModel of specificModels) {
+        if (model.toLowerCase() === specificModel) {
           return true;
         }
       }
 
       const supportedContains = ["llama-3.1"];
-      for (const model of supportedContains) {
-        if (model.toLowerCase().includes(model)) {
+      for (const contained of supportedContains) {
+        if (model.toLowerCase().includes(contained)) {
           return true;
         }
       }
 
       return false;
     },
+    zAI: (model) => {
+      const lower = model.toLowerCase();
+      return lower.startsWith("glm-4") || lower.startsWith("glm-5");
+    },
     moonshot: (model) => {
       // support moonshot models
       // https://platform.moonshot.ai/docs/pricing/chat#concepts
@@ -426,7 +441,9 @@ export function isRecommendedAgentModel(modelName: string): boolean {
     [/deepseek/, /r1|reasoner/],
     [/gemini/, /2\.5/, /pro/],
     [/gemini/, /3-pro/],
-    [/gpt/, /-5|5\.1/],
+    [/gpt/, /-5|5\.1|5\.2/],
+    [/gpt-4\.1/],
+    [/codex/],
     [/claude/, /sonnet/, /3\.7|3-7|-4/],
     [/claude/, /opus/, /-4/],
     [/grok-code/],

docs/customize/model-providers/more/moonshot.mdx

@@ -1,9 +1,9 @@
 ---
 title: "Moonshot AI"
-description: "Configure Moonshot AI's language models with Continue, including setup instructions for their 8K, 32K, and 128K context window models with competitive pricing"
+description: "Configure Moonshot AI's language models with Continue, including Kimi K2, Kimi K2.5, and Moonshot v1 models with competitive pricing"
 ---
 
-[Moonshot AI](https://platform.moonshot.cn/) provides high-quality large language model services with competitive pricing and excellent performance.
+[Moonshot AI](https://platform.moonshot.cn/) provides high-quality large language model services with competitive pricing and excellent performance, including the Kimi series of models.
 
 ## Configuration
 
@@ -20,20 +20,20 @@ To use Moonshot AI models, you need to:
   schema: v1
 
   models:
-    - name: Moonshot
+    - name: Kimi K2
       provider: moonshot
-      model: moonshot-v1-8k
+      model: kimi-k2
       apiKey: <YOUR_MOONSHOT_API_KEY>
   ```
   </Tab>
-  <Tab title="JSON">
+  <Tab title="JSON (Deprecated)">
   ```json title="config.json"
   {
     "models": [
       {
-        "title": "Moonshot",
+        "title": "Kimi K2",
         "provider": "moonshot",
-        "model": "moonshot-v1-8k",
+        "model": "kimi-k2",
         "apiKey": "<YOUR_MOONSHOT_API_KEY>"
       }
     ]
@@ -46,6 +46,8 @@ To use Moonshot AI models, you need to:
 
 Moonshot AI currently provides the following models:
 
+- `kimi-k2`: Kimi K2, a powerful MoE model with 128K context window
+- `kimi-k2.5`: Kimi K2.5, the latest generation with enhanced reasoning and 128K context window
 - `moonshot-v1-8k`: Base model with 8K context window
 - `moonshot-v1-32k`: Base model with 32K context window
 - `moonshot-v1-128k`: Base model with 128K context window
@@ -70,24 +72,24 @@ Here's a complete configuration example:
   schema: v1
 
   models:
-    - name: Moonshot-8K
+    - name: Kimi K2
       provider: moonshot
-      model: moonshot-v1-8k
+      model: kimi-k2
       apiKey: <YOUR_MOONSHOT_API_KEY>
       defaultCompletionOptions:
         temperature: 0.7
         topP: 0.95
         maxTokens: 2048
   ```
   </Tab>
-  <Tab title="JSON">
+  <Tab title="JSON (Deprecated)">
   ```json title="config.json"
   {
     "models": [
       {
-        "title": "Moonshot-8K",
+        "title": "Kimi K2",
         "provider": "moonshot",
-        "model": "moonshot-v1-8k",
+        "model": "kimi-k2",
         "apiKey": "<YOUR_MOONSHOT_API_KEY>",
         "completionOptions": {
           "temperature": 0.7,

docs/customize/model-providers/more/zai.mdx

@@ -0,0 +1,59 @@
+---
+title: "Z.ai"
+description: "Configure Z.ai's GLM models with Continue, including GLM-5, GLM-4.7, and GLM-4.5"
+---
+
+[Z.ai](https://z.ai/) (formerly Zhipu AI) provides the GLM family of large language models with strong multilingual capabilities.
+
+<Info>Get an API key from the [Z.ai Platform](https://z.ai/manage-apikey/apikey-list)</Info>
+
+## Configuration
+
+<Tabs>
+  <Tab title="YAML">
+  ```yaml title="config.yaml"
+  name: My Config
+  version: 0.0.1
+  schema: v1
+
+  models:
+    - name: GLM-5
+      provider: zAI
+      model: glm-5
+      apiKey: <YOUR_ZAI_API_KEY>
+  ```
+  </Tab>
+  <Tab title="JSON (Deprecated)">
+  ```json title="config.json"
+  {
+    "models": [
+      {
+        "title": "GLM-5",
+        "provider": "zAI",
+        "model": "glm-5",
+        "apiKey": "<YOUR_ZAI_API_KEY>"
+      }
+    ]
+  }
+  ```
+  </Tab>
+</Tabs>
+
+## Available Models
+
+- `glm-5`: Flagship model with 128K context window
+- `glm-4.7`: Latest generation model with 128K context window
+- `glm-4-plus`: Enhanced GLM-4 model with 128K context window
+- `glm-4.5`: GLM-4.5 model with 128K context window
+
+## Configuration Options
+
+| Option    | Description          | Default                                |
+| --------- | -------------------- | -------------------------------------- |
+| `apiKey`  | Z.ai API key         | Required                               |
+| `apiBase` | API base URL         | `https://api.z.ai/api/paas/v4/`       |
+| `model`   | Model name to use    | `glm-5`                               |
+
+<Tip>
+Z.ai offers a dedicated [Coding Plan](https://docs.z.ai/api-reference/introduction) with a subscription-based endpoint optimized for coding tasks. To use it, set `apiBase` to `https://api.z.ai/api/coding/paas/v4/`. This endpoint uses prompt-based pricing rather than token-based pricing and may be more cost-effective for coding workflows.
+</Tip>

docs/docs.json

@@ -176,10 +176,12 @@
                       "customize/model-providers/more/llamastack",
                       "customize/model-providers/more/mimo",
                       "customize/model-providers/more/mistral",
+                      "customize/model-providers/more/moonshot",
                       "customize/model-providers/more/nous",
                       "customize/model-providers/more/nvidia",
                       "customize/model-providers/more/together",
-                      "customize/model-providers/more/xAI"
+                      "customize/model-providers/more/xAI",
+                      "customize/model-providers/more/zai"
                     ]
                   }
                 ]