GLM 5.3 Flash

GLM 5.3 Flash (z-ai/glm-5.3-flash) is a fast multimodal model from Z.ai: efficient coding and long-horizon agent tasks. Its hybrid sparse and linear attention architecture maintains accurate long-context behavior (1M tokens) while cutting compute overhead. It accepts text, images and video as input.

At 8.4 ₽ per 1M input tokens it is one of the cheapest paid text models in the catalog — a workhorse for subagents and bulk runs where the count goes into thousands of requests sharing the same prompt: the repeating prefix is read from cache at 1.7 ₽.

Parameters and capabilities

Parameter	Value
model	z-ai/glm-5.3-flash (or the short glm-5.3-flash)
Context	1M tokens
Reasoning	adjustable, can be disabled (off)
reasoning_effort	low, medium, high
Image input	yes — image_url parts with data: URLs
File input	no — a file part returns 400
Cache read	1.7 ₽ per 1M tokens
Cache write	not published in the catalog

Video input is listed in the model’s description, but /v1/chat/completions accepts only images from multimodal parts: the first image_url part of the last user message goes to the model, and external HTTP links are not supported — data:image/...;base64,... only.

Examples

Bulk processing

Python

from openai import OpenAI

client = OpenAI(
    base_url="https://api.mixen.ai/v1",
    api_key="mxn-...",
)

resp = client.chat.completions.create(
    model="z-ai/glm-5.3-flash",
    messages=[{
        "role": "user",
        "content": "Classify this ticket as: billing, technical, other. Answer in one line. Text: ...",
    }],
)
print(resp.choices[0].message.content)

cURL

curl https://api.mixen.ai/v1/chat/completions \
  -H "Authorization: Bearer $MIXEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "z-ai/glm-5.3-flash",
    "messages": [{
      "role": "user",
      "content": "Classify this ticket as: billing, technical, other. Answer in one line. Text: ..."
    }]
  }'

An image on input

The model sees images — pass them as data: URLs inside the message:

{
  "model": "z-ai/glm-5.3-flash",
  "messages": [{
    "role": "user",
    "content": [
      {"type": "text", "text": "What is in the screenshot and which error is visible?"},
      {"type": "image_url", "image_url": {"url": "data:image/png;base64,iVBORw0KG..."}}
    ]
  }]
}

For a pipeline sharing one system prompt, context caching kicks in automatically — a stable prefix makes such runs even cheaper.

Price

Per 1M tokens: input 8.4 ₽ ($0.094), output 27.9 ₽ ($0.312), cache read 1.7 ₽ ($0.019); a separate cache-write price is not published by the catalog. Current prices — catalog.

Strengths and limits

• Efficient coding and long-horizon agent tasks at one of the lowest input prices in the catalog — bulk runs and subagents do not eat the budget.
• 1M-token context without accuracy loss on long inputs — as claimed by the model’s architecture.
• Multimodal input: text and images in one request.
• Reasoning is adjustable: low/medium/high plus an honest off for simple tasks.
• Files are not accepted as input — file parts return 400.
• Video input is a stated capability of the model, but over /v1 only images can be passed.

All models — in the catalog. The general text-model workflow — chat guide.