GLM 4.6V Flash

zhipuzhipu/glm-4.6v-flash

API guide

Vision chat completion (image / video / PDF input)

GLM-4.6V-Flash is accessible through the standard OpenAI /v1/chat/completions endpoint at api.therouter.ai. Pass images as image_url content blocks, exactly as you would with GPT-4o. The model supports text, image, video, and PDF inputs in a single request. Function Calling works via the standard tools / tool_choice parameters — images and visual tool outputs are handled natively. The API is free; no credits are consumed.

cURL

# Image understanding — free via TheRouter
curl https://api.therouter.ai/v1/chat/completions \
  -H "Authorization: Bearer $THEROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "zhipu/glm-4.6v-flash",
    "messages": [
      {
        "role": "user",
        "content": [
          {
            "type": "image_url",
            "image_url": { "url": "https://therouter.ai/assets/vision-sample.png" }
          },
          {
            "type": "text",
            "text": "Summarise the key trends shown in this chart."
          }
        ]
      }
    ],
    "max_tokens": 1024
  }'

Multimodal Function Calling

GLM-4.6V-Flash is the first GLM vision model with native multimodal Function Calling. Define tools in the standard OpenAI tools format; the model can pass images directly as tool arguments and interpret visual tool outputs (charts, screenshots, search results) without converting them to text first.

cURL

curl https://api.therouter.ai/v1/chat/completions \
  -H "Authorization: Bearer $THEROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "zhipu/glm-4.6v-flash",
    "messages": [
      {
        "role": "user",
        "content": [
          { "type": "image_url", "image_url": { "url": "https://therouter.ai/assets/vision-sample.png" } },
          { "type": "text", "text": "Is the server status healthy? Trigger an alert if not." }
        ]
      }
    ],
    "tools": [{
      "type": "function",
      "function": {
        "name": "trigger_alert",
        "description": "Trigger an ops alert",
        "parameters": {
          "type": "object",
          "properties": {
            "service_name": { "type": "string" },
            "severity": { "type": "string", "enum": ["low","medium","high"] },
            "message": { "type": "string" }
          },
          "required": ["service_name","severity","message"]
        }
      }
    }],
    "tool_choice": "auto",
    "max_tokens": 512
  }'

Fact ledger — every claim on this page traces here

source	URL	retrieved
Release date	huggingface.co ↗	2026-06-09	verified
Parameter count	huggingface.co ↗	2026-06-09	verified
Architecture	huggingface.co ↗	2026-06-09	verified
Context window	docs.z.ai ↗	2026-06-09	verified
Max output tokens	huggingface.co ↗	2026-06-09	verified
Input modalities	docs.z.ai ↗	2026-06-09	verified
Output modality	docs.z.ai ↗	2026-06-09	verified
Pricing	docs.z.ai ↗	2026-06-09	verified
Native Function Calling	huggingface.co ↗	2026-06-09	verified
Training cutoff	—	—	unknown
License	huggingface.co ↗	2026-06-09	to verify
Local deployment support	huggingface.co ↗	2026-06-09	verified
Recommended decoding parameters	huggingface.co ↗	2026-06-09	verified
Z.ai releases GLM-4.6V series with native multimodal Function Calling	huggingface.co ↗	2026-06-09	verified