GLM-5.3-Flash

zai/glm-5.3-flash is a Z.AI model that processes images alongside code and text and supports tool calling. It has a 1,000,000-token context window and a 131,072-token output ceiling per response.

Model	Context	Output ceiling	Inputs
zai/glm-5.3-flash	1,000,000	131,072	Text, image

Audio, video, and document input are not advertised. Convert the pages you need from a PDF or office document to images instead of sending the document directly.

Pricing

Prices are per one million tokens. The two discounts are not stacked.

Period	Input	Cache read	Output
Z.AI promotion through September 9, 2026 at 16:00 UTC	$0.075	$0.015	$0.250
LLMTR agreement after September 9, 2026 at 16:00 UTC	$0.135	$0.027	$0.450
Undiscounted Z.AI API list price	$0.150	$0.030	$0.500

The exact promotion cutoff is 2026-09-09T16:00:00.000Z. At that instant, LLMTR's agreement price—10% below the Z.AI list price—takes effect automatically.

Text request

curl "$LLMTR_BASE_URL/v1/chat/completions" \
  -H "Authorization: Bearer llmtr-your_key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "zai/glm-5.3-flash",
    "messages": [
      { "role": "user", "content": "Suggest three improvements to this TypeScript function." }
    ],
    "max_tokens": 256
  }'

Reasoning

Reasoning is always on for this model and cannot be disabled. Plain zai/glm-5.3-flash requests run at low. For harder work, select only low, high, or max through a model suffix or the reasoning_effort field:

{
  "model": "zai/glm-5.3-flash",
  "reasoning_effort": "high",
  "messages": [{ "role": "user", "content": "Trace this stack trace to its root cause." }],
  "max_tokens": 512
}

:fast, reasoning: false, and thinking: { "type": "disabled" } are rejected. See Z.AI Thinking Control for the shared behavior.

Image input

Send image_url in an OpenAI-compatible content array:

{
  "model": "zai/glm-5.3-flash",
  "messages": [{
    "role": "user",
    "content": [
      { "type": "text", "text": "List the accessibility issues in this interface." },
      { "type": "image_url", "image_url": { "url": "https://example.com/screenshot.png" } }
    ]
  }],
  "max_tokens": 256
}

Tool calling and cache

The model supports function calling through the tools and tool_choice fields. When exactly one function must be selected, send only that tool and use tool_choice: "required". See Tool Calling for tool schemas and result messages.

Input tokens that hit Z.AI's cache use the separate CACHE_READ rate. When the usage response reports the cache metric, LLMTR bills those tokens at the lower cache-read price above; the remaining input tokens stay on the regular input rate.

Sources for the model and pricing limits: Z.AI model guide, core parameters, and pricing.