GLM 5.3 Prime

GLM 5.3 Prime (z-ai/glm-5.3-prime) — the top tier of Z.ai’s GLM 5.3 line: reasons deeper and stays more stable through long agentic sessions than the base GLM 5.3. Thinking is always on; the depth is low/high/max. Text only.

Parameters and capabilities

Parameter	Value
model	z-ai/glm-5.3-prime (short alias glm-5.3-prime)
reasoning_effort	low, high, max
Disabling reasoning	no — the model always thinks
Images on input	no — an image_url part returns 400
Files on input	no — a file part returns 400
Cache read	$0.7 per 1M tokens

Example

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-prime",
    "reasoning_effort": "high",
    "messages": [{"role": "user", "content": "Design a fault-tolerant ingestion pipeline for 10k events/s"}]
  }'

Pricing

Per 1M tokens: input $3.5, output $11, cache read $0.7. Current prices — catalog.

Strengths and limitations

• Reasoning depth low/high/max; off and medium are not supported by the model.
• Reasoning is always on: its tokens are billed on every request.
• Text only: images and files return 400.
• The base GLM 5.3 with a 1M context — its page.

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