feat(compute): add GLM-5.3 Flash to Coding Plan (#83)

- add GLM-5.3 Flash to the Z.AI Coding Plan preset
- declare multimodal and always-on reasoning capabilities
- add regression coverage and bump preset data version

__tests__/validate.test.mjs

@@ -378,6 +378,28 @@ describe('真实数据全量校验', () => {
     assert.equal(modelSpec.routing.reasoning.defaultMode, 'xhigh')
   })
 
+  it('GLM-5.3-Flash 应在 Coding Plan 中提供强制思考的多模态接入', () => {
+    const codingPlan = JSON.parse(readFileSync(join(ROOT, 'compute', 'coding-plans', 'zhipu-coding.json'), 'utf8'))
+    const specFile = JSON.parse(readFileSync(join(ROOT, 'compute', 'model-specs', 'zhipu.json'), 'utf8'))
+    const modelId = 'glm-5.3-flash'
+    const model = codingPlan.models.find((item) => item.modelName === modelId)
+    const modelSpec = specFile.specs.find((item) => item.id === modelId)
+
+    assert.ok(model, `智谱 Coding Plan 缺少 ${modelId}`)
+    assert.equal(model.contextWindow, 1048576)
+    assert.equal(model.maxOutputTokens, 131072)
+    assert.ok(model.capabilities.includes('vision'))
+    assert.ok(model.capabilities.includes('video_understanding'))
+    assert.deepEqual(model.serviceType, ['chat', 'reasoning', 'vision'])
+    assert.equal(model.extra.thinkingOnly, true)
+    assert.deepEqual(model.extra.reasoning.supportedEfforts, ['low', 'high', 'max'])
+    assert.equal(model.extra.reasoning.defaultEffort, 'max')
+
+    assert.ok(modelSpec, `model-specs 缺少 ${modelId}`)
+    assert.equal(modelSpec.spec.extra.thinkingOnly, true)
+    assert.deepEqual(modelSpec.routing.reasoning.supportedModes, ['auto', 'low', 'high', 'max'])
+  })
+
   it('基础模型与后缀版本应使用独立精确规格，避免互相误匹配', () => {
     const qwen = JSON.parse(readFileSync(join(ROOT, 'compute', 'model-specs', 'qwen.json'), 'utf8'))
     const deepseek = JSON.parse(readFileSync(join(ROOT, 'compute', 'model-specs', 'deepseek.json'), 'utf8'))

compute/coding-plans/zhipu-coding.json

@@ -27,6 +27,48 @@
     "glm-4.5-air"
   ],
   "models": [
+    {
+      "modelName": "glm-5.3-flash",
+      "displayName": "GLM-5.3-Flash (Coding)",
+      "serviceType": [
+        "chat",
+        "reasoning",
+        "vision"
+      ],
+      "description": "智谱 GLM-5.3-Flash Coding Plan 版本，原生多模态高效模型，1M 上下文，128K 最大输出，强制开启思考",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "vision",
+        "image_understanding",
+        "video_understanding",
+        "long_context",
+        "tool_use",
+        "agent",
+        "multilingual",
+        "fast"
+      ],
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "supportsThinking": true,
+        "thinkingDefault": true,
+        "thinkingOnly": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "max"
+        }
+      },
+      "source": "preset"
+    },
     {
       "modelName": "glm-5.2",
       "apiModelId": "glm-5.2[1m]",

manifest.json

@@ -1,6 +1,6 @@
 {
   "version": "1.0.0",
-  "presetDataVersion": 98,
+  "presetDataVersion": 99,
-  "updatedAt": "2026-08-27",
+  "updatedAt": "2026-08-28",
   "description": "DesireCore 官方配置中心"
 }