feat(zai): add GLM-5.3-Flash Coding Plan support

.env.example

@@ -192,8 +192,13 @@ ANTHROPIC_API_KEY=sk-ant-your-key-here
 # For Z.AI GLM Coding Plan, set:
 # OPENAI_BASE_URL=https://api.z.ai/api/coding/paas/v4
 # OPENAI_MODEL=glm-5.2
-# Optional: OPENAI_MODEL=glm-5.3 (the default remains glm-5.2)
+# Optional: OPENAI_MODEL=glm-5.3-flash (supports image input; the default remains glm-5.2)
+# Optional: OPENAI_MODEL=glm-5.3
 # Optional: OPENAI_MODEL=GLM-5.1, GLM-5-Turbo, GLM-4.7, or GLM-4.5-Air
+# Optional GLM-5.3-Flash thinking controls:
+# OPENAI_MODEL='glm-5.3-flash?reasoning=low'   # requests Z.AI reasoning_effort=low
+# OPENAI_MODEL='glm-5.3-flash?reasoning=high'  # requests Z.AI reasoning_effort=high
+# OPENAI_MODEL='glm-5.3-flash?reasoning=xhigh' # maps to Z.AI reasoning_effort=max
 # Optional GLM-5.3 thinking controls:
 # OPENAI_MODEL='glm-5.3?reasoning=low'      # requests Z.AI reasoning_effort=low
 # OPENAI_MODEL='glm-5.3?reasoning=high'     # requests Z.AI reasoning_effort=high

README.md

@@ -306,7 +306,7 @@ Advanced and source-build guides:
 | Provider | Setup Path | Notes |
 | --- | --- | --- |
 | OpenAI-compatible | `/provider` or env vars | Works with OpenAI, OpenRouter, DeepSeek, Groq, Mistral, LM Studio, and other compatible `/v1` servers |
-| Z.AI GLM Coding Plan | `/provider` or OpenAI-compatible env vars | Uses `OPENAI_API_KEY` at `https://api.z.ai/api/coding/paas/v4` and defaults to `glm-5.2` |
+| Z.AI GLM Coding Plan | `/provider` or OpenAI-compatible env vars | Uses `OPENAI_API_KEY` at `https://api.z.ai/api/coding/paas/v4`, defaults to `glm-5.2`, and offers the vision-capable `glm-5.3-flash` option |
 | AI/ML API | `/provider` or `AIMLAPI_API_KEY` ([setup guide](docs/aimlapi-setup.md)) | Uses `https://api.aimlapi.com/v1`, auto-detects the OpenAI-compatible route from `AIMLAPI_API_KEY`, sends OpenClaude attribution headers, and discovers chat-capable models from the public `/models` catalog |
 | Concentrate | `/provider` or `CONCENTRATE_API_KEY` | Unified OpenAI-compatible gateway at `https://api.concentrate.ai/v1`; defaults to `deepseek-v4-flash` and auto-discovers the chat model catalog |
 | LLMTR | `/provider` or OpenAI-compatible env vars | Multi-model gateway at `https://llmtr.com/v1`; `/provider` and `--provider llmtr` default to `deepseek/deepseek-v4-flash`, while raw env setup must set `OPENAI_BASE_URL=https://llmtr.com/v1` and `OPENAI_MODEL`; accepts `LLMTR_API_KEY` or `OPENAI_API_KEY` after the route is selected and discovers tool-capable Chat Completions models from the public catalog |
@@ -373,7 +373,7 @@ OpenClaude supports multiple providers, but behavior is not identical across all
 - Some providers impose lower output caps than the CLI defaults, and OpenClaude adapts where possible
 - AI/ML API uses the OpenAI-compatible route, defaults to `gpt-4o`, and only surfaces chat-capable models from its public catalog
 - Gitlawb Opengateway is the fresh-install startup default and requires an API key from https://gitlawb.com/opengateway/keys. It uses one OpenAI-compatible base URL; switch between `mimo-*` and `google/gemini-3.1-flash-lite-preview` with `/model`, and do not pin the base URL to `/v1/xiaomi-mimo`.
-- Z.AI GLM Coding Plan uses `https://api.z.ai/api/coding/paas/v4` with `glm-5.2` by default. GLM-5.3 is selectable as `glm-5.3`; use `glm-5.3?reasoning=low`, `glm-5.3?reasoning=high`, or `glm-5.3?reasoning=xhigh` to request its documented low, high, or maximum effort. The existing GLM-5.2 query controls remain supported.
+- Z.AI GLM Coding Plan uses `https://api.z.ai/api/coding/paas/v4` with `glm-5.2` by default. On that direct route, GLM-5.3-Flash is selectable as `glm-5.3-flash` and accepts image input. Its OpenClaude effort choices are `low`, `high`, and `xhigh`, with `xhigh` requesting Z.AI `max`; GLM-5.3 and the existing GLM-5.2 query controls remain supported. These capabilities are not claimed for gateways that happen to carry the same model name.
 - Xiaomi MiMo uses `api-key` header auth on the direct OpenAI-compatible route and currently does not support `/usage` reporting in OpenClaude
 - GitHub Copilot serializes sub-agent execution by default to reduce Premium Request consumption — see [Agent Routing and Step Limits](docs/agent-routing.md#github-copilot-sub-agent-optimization) for tuning
 

docs/integrations/reasoning-effort.md

@@ -46,7 +46,9 @@ Use `mode: 'always-on'` with `wireFormat: 'none'` for models that emit reasoning
 
 Currently wired metadata formats are `reasoning_effort`, `deepseek_compatible`, and `zai_compatible`. The descriptor type also reserves `reasoning_object` and `thinking_type`, but those formats are not request-plumbed yet and should not be used to enable `/effort`.
 
-For `deepseek_compatible` and `zai_compatible`, metadata levels must be limited to `high` and/or `xhigh`. These serializers emit provider `high` for `high` and provider `max` for `xhigh`; they cannot faithfully represent `low`, `medium`, or standard `max` as distinct UI levels.
+For `deepseek_compatible`, metadata levels must be limited to `high` and/or `xhigh`. The serializer emits provider `high` for OpenClaude `high` and provider `max` for OpenClaude `xhigh`; it cannot represent `low`, `medium`, or standard `max` as distinct levels.
+
+For `zai_compatible`, exact route/model entries may expose verified `low`, `high`, and `xhigh` levels. These serialize as provider `low`, `high`, and `max`, respectively. `medium` is not a distinct GLM-5.3-Flash level, and standard OpenClaude `max` is not added as another picker level. Disabling thinking must be verified separately; low effort is not equivalent to no reasoning.
 
 ## Adding Support
 

src/integrations/brands/glm.ts

@@ -13,6 +13,7 @@ export default defineBrand({
     supportsPreciseTokenCount: false,
   },
   modelIds: [
+    'glm-5.3-flash',
     'glm-5.3',
     'glm-5.2',
     'GLM-5.1',

src/integrations/models/glm.ts

@@ -32,6 +32,21 @@ function glmModel(
 }
 
 export default [
+  defineModel({
+    id: 'glm-5.3-flash',
+    label: 'GLM 5.3 Flash',
+    brandId: 'glm',
+    vendorId: 'zai',
+    classification: ['chat', 'reasoning', 'vision', 'coding'],
+    defaultModel: 'glm-5.3-flash',
+    capabilities: {
+      ...glmCapabilities,
+      supportsVision: true,
+    },
+    contextWindow: 1_000_000,
+    maxOutputTokens: 131_072,
+    runtimeMetadataScope: 'catalog',
+  }),
   glmModel('glm-5.3', 'GLM 5.3', 1_000_000, 131_072, 'catalog'),
   defineModel({
     id: 'glm-5v-turbo',

src/integrations/runtimeMetadata.test.ts

@@ -16,6 +16,7 @@ import {
   getDiscoveryCacheKey,
   getRouteDiscoveryHeaders,
 } from './discoveryService'
+import { resolveActiveRouteIdFromEnv } from './routeMetadata.js'
 import { setClaudeConfigHomeDirForTesting } from '../utils/envUtils.js'
 import glmBrand from './brands/glm.js'
 import glmModels from './models/glm.js'
@@ -24,6 +25,53 @@ import zaiVendor from './vendors/zai.js'
 const originalConfigDir = process.env.CLAUDE_CONFIG_DIR
 
 describe('Z.AI GLM-5.3 descriptor contract', () => {
+  it('wires the verified GLM-5.3-Flash descriptor and direct catalog contract', () => {
+    const model = glmModels.find(candidate => candidate.id === 'glm-5.3-flash')
+    expect(model).toMatchObject({
+      id: 'glm-5.3-flash',
+      label: 'GLM 5.3 Flash',
+      vendorId: 'zai',
+      brandId: 'glm',
+      classification: ['chat', 'reasoning', 'vision', 'coding'],
+      defaultModel: 'glm-5.3-flash',
+      contextWindow: 1_000_000,
+      maxOutputTokens: 131_072,
+      runtimeMetadataScope: 'catalog',
+      capabilities: {
+        supportsVision: true,
+        supportsStreaming: true,
+        supportsFunctionCalling: true,
+        supportsJsonMode: true,
+        supportsReasoning: true,
+        supportsPreciseTokenCount: false,
+      },
+    })
+
+    expect(glmBrand.modelIds?.[0]).toBe('glm-5.3-flash')
+    expect(glmBrand.modelIds?.[1]).toBe('glm-5.3')
+
+    const catalogEntries = zaiVendor.catalog?.models ?? []
+    const flashEntries = catalogEntries.filter(entry => entry.id === 'glm-5.3-flash')
+    expect(flashEntries).toHaveLength(1)
+    expect(catalogEntries[0]).toMatchObject({
+      id: 'glm-5.3-flash',
+      apiName: 'glm-5.3-flash',
+      label: 'GLM-5.3-Flash',
+      modelDescriptorId: 'glm-5.3-flash',
+      reasoning: {
+        mode: 'levels',
+        levels: ['low', 'high', 'xhigh'],
+        wireFormat: 'zai_compatible',
+      },
+      transportOverrides: {
+        openaiShim: { enableToolStreaming: true },
+      },
+    })
+    expect(catalogEntries[1]?.id).toBe('glm-5.3')
+    expect(zaiVendor.catalog?.source).toBe('static')
+    expect(zaiVendor.defaultModel).toBe('glm-5.2')
+  })
+
   it('wires the verified shared model, brand, and direct catalog entry without changing the default', () => {
     const model = glmModels.find(candidate => candidate.id === 'glm-5.3')
     expect(model).toMatchObject({
@@ -45,9 +93,9 @@ describe('Z.AI GLM-5.3 descriptor contract', () => {
         supportsPreciseTokenCount: false,
       },
     })
-    expect(glmBrand.modelIds?.[0]).toBe('glm-5.3')
+    expect(glmBrand.modelIds).toContain('glm-5.3')
 
-    const catalogEntry = zaiVendor.catalog?.models?.[0]
+    const catalogEntry = zaiVendor.catalog?.models?.find(entry => entry.id === 'glm-5.3')
     expect(catalogEntry).toMatchObject({
       id: 'glm-5.3',
       apiName: 'glm-5.3',
@@ -166,6 +214,25 @@ describe('resolveModelRuntimeLimits', () => {
       }
     })
   })
+  it.each([
+    'glm-5.3-flash',
+    'glm-5.3-flash?reasoning=low',
+    'glm-5.3-flash?reasoning=high',
+    'glm-5.3-flash?reasoning=xhigh',
+    'glm-5.3-flash?thinking=disabled',
+  ])('uses verified Z.AI GLM-5.3-Flash runtime limits for %s', model => {
+    const limits = resolveModelRuntimeLimits({
+      model,
+      processEnv: {
+        CLAUDE_CODE_USE_OPENAI: '1',
+        OPENAI_BASE_URL: 'https://api.z.ai/api/coding/paas/v4',
+      },
+    })
+
+    expect(limits.contextWindow).toBe(1_000_000)
+    expect(limits.maxOutputTokens).toBe(131_072)
+  })
+
   it.each([
     'glm-5.3',
     'glm-5.3?reasoning=low',
@@ -197,16 +264,37 @@ describe('resolveModelRuntimeLimits', () => {
   })
 
   it.each([
-    ['NVIDIA NIM', 'https://integrate.api.nvidia.com/v1', { NVIDIA_NIM: '1' }],
-    ['OpenRouter', 'https://openrouter.ai/api/v1', { CLAUDE_CODE_USE_OPENAI: '1' }],
-    ['custom endpoint', 'https://proxy.example.test/v1', { CLAUDE_CODE_USE_OPENAI: '1' }],
-  ] as const)('does not leak direct Z.AI GLM-5.3 limits onto %s', (_name, baseUrl, routeEnv) => {
+    ['NVIDIA NIM', 'https://integrate.api.nvidia.com/v1', { CLAUDE_CODE_USE_OPENAI: '1', NVIDIA_NIM: '1' }, 'nvidia-nim'],
+    ['OpenRouter', 'https://openrouter.ai/api/v1', { CLAUDE_CODE_USE_OPENAI: '1' }, 'openrouter'],
+    ['custom endpoint', 'https://proxy.example.test/v1', { CLAUDE_CODE_USE_OPENAI: '1' }, 'custom'],
+  ] as const)('does not leak direct Z.AI GLM-5.3-Flash limits onto %s', (_name, baseUrl, routeEnv, expectedRoute) => {
+    const processEnv = {
+      ...routeEnv,
+      OPENAI_BASE_URL: baseUrl,
+    }
+    expect(resolveActiveRouteIdFromEnv(processEnv)).toBe(expectedRoute)
+    expect(resolveModelRuntimeLimits({
+      model: 'glm-5.3-flash',
+      processEnv,
+    })).toEqual({
+      contextWindow: undefined,
+      maxOutputTokens: undefined,
+    })
+  })
+
+  it.each([
+    ['NVIDIA NIM', 'https://integrate.api.nvidia.com/v1', { CLAUDE_CODE_USE_OPENAI: '1', NVIDIA_NIM: '1' }, 'nvidia-nim'],
+    ['OpenRouter', 'https://openrouter.ai/api/v1', { CLAUDE_CODE_USE_OPENAI: '1' }, 'openrouter'],
+    ['custom endpoint', 'https://proxy.example.test/v1', { CLAUDE_CODE_USE_OPENAI: '1' }, 'custom'],
+  ] as const)('does not leak direct Z.AI GLM-5.3 limits onto %s', (_name, baseUrl, routeEnv, expectedRoute) => {
+    const processEnv = {
+      ...routeEnv,
+      OPENAI_BASE_URL: baseUrl,
+    }
+    expect(resolveActiveRouteIdFromEnv(processEnv)).toBe(expectedRoute)
     expect(resolveModelRuntimeLimits({
       model: 'glm-5.3',
-      processEnv: {
-        ...routeEnv,
-        OPENAI_BASE_URL: baseUrl,
-      },
+      processEnv,
     })).toEqual({
       contextWindow: undefined,
       maxOutputTokens: undefined,
@@ -431,6 +519,29 @@ describe('resolveOpenAIShimRuntimeContext - Z.A.I GLM-5.3', () => {
   })
 })
 
+describe('resolveOpenAIShimRuntimeContext - Z.A.I GLM-5.3-Flash', () => {
+  it.each([
+    'glm-5.3-flash',
+    'glm-5.3-flash?reasoning=xhigh',
+    'glm-5.3-flash?thinking=disabled',
+  ])('uses the explicit direct-route GLM-5.3-Flash contract for %s', model => {
+    const result = resolveOpenAIShimRuntimeContext({
+      model,
+      baseUrl: 'https://api.z.ai/api/coding/paas/v4',
+      processEnv: {},
+    })
+
+    expect(result.routeId).toBe('zai')
+    expect(result.catalogEntry?.id).toBe('glm-5.3-flash')
+    expect(result.openaiShimConfig.thinkingRequestFormat).toBe('zai-compatible')
+    expect(result.openaiShimConfig.preserveReasoningContent).toBe(true)
+    expect(result.openaiShimConfig.requireReasoningContentOnAssistantMessages).toBe(true)
+    expect(result.openaiShimConfig.maxTokensField).toBe('max_tokens')
+    expect(result.openaiShimConfig.removeBodyFields).toContain('store')
+    expect(result.openaiShimConfig.enableToolStreaming).toBe(true)
+  })
+})
+
 describe('resolveOpenAIShimRuntimeContext - GLM on a non-Z.AI gateway (#1896)', () => {
   it('infers the GLM reasoning shim but not tool streaming for a third-party gateway', () => {
     const result = resolveOpenAIShimRuntimeContext({

src/integrations/vendors/zai.ts

@@ -44,6 +44,22 @@ export default defineVendor({
   catalog: {
     source: 'static',
     models: [
+      {
+        id: 'glm-5.3-flash',
+        apiName: 'glm-5.3-flash',
+        label: 'GLM-5.3-Flash',
+        modelDescriptorId: 'glm-5.3-flash',
+        reasoning: {
+          mode: 'levels',
+          levels: ['low', 'high', 'xhigh'],
+          wireFormat: 'zai_compatible',
+        },
+        transportOverrides: {
+          openaiShim: {
+            enableToolStreaming: true,
+          },
+        },
+      },
       {
         id: 'glm-5.3',
         apiName: 'glm-5.3',