GLM-5.3 API: Pricing, Flash Rate and Setup

GLM-5.3 is Zhipu's current flagship, live as model ID glm-5.3. The list price is $1.40 per 1M input tokens and $4.40 per 1M output tokens — the same sticker as GLM-5.2. The cheap sibling is glm-5.3-flash at $0.15 / $0.50, with vision.

The number most price tables will get wrong this week is the Flash promo. Z.ai advertised $0.075 / $0.25 through 9 September 2026. RouterPlex bills the standard published list rate of $0.15 / $0.50.

GLM-5.3 API pricing #

All figures are USD per 1M tokens.

A $5 prepaid balance on GLM-5.3 buys roughly 3.6 million input tokens, or 1.1 million output tokens, if a workload used only one category. On Flash that is about 33 million input tokens, or 10 million output tokens.

Cached prompt tokens on this catalog are $0.26 per 1M on GLM-5.3 and $0.03 per 1M on GLM-5.3 Flash.

The Flash promo is not this bill #

Z.ai ran a launch promotion at $0.075 input / $0.25 output per 1M through 9 September 2026. That is half the standard list rate. RouterPlex does not pass that window through. Budget Flash at $0.15 / $0.50.

If a comparison table or a screenshot from the lab console shows $0.075, it is the promo, not this route.

Cost per agent turn #

One 40,000-token prompt returning 2,000 tokens, no cache:

The arithmetic for GLM-5.3: 40,000 × $1.40 / 1M + 2,000 × $4.40 / 1M = $0.056 + $0.0088 = $0.0648. Flash is 40,000 × $0.15 / 1M + 2,000 × $0.50 / 1M = $0.006 + $0.001 = $0.007.

GLM-5.3 is priced as a mid-flagship, not a volume model. Flash is the volume GLM, and it is the one with image input.

Live pages: GLM-5.3, GLM-5.3 Flash, GLM-5.2.

GLM-5.3 specifications #

Reasoning tokens are billed as output. A short reply is not necessarily a cheap one — check usage.completion_tokens_details.reasoning_tokens if a simple prompt costs more than you expected.

Call GLM-5.3 with RouterPlex #

curl https://api.routerplex.com/v1/chat/completions \
  -H "Authorization: Bearer $ROUTERPLEX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5.3",
    "messages": [
      {"role": "user", "content": "Name the riskiest assumption in this migration plan."}
    ]
  }'

Python uses the regular OpenAI client — only the base URL, key and model ID change:

import os
from openai import OpenAI
 
client = OpenAI(
    api_key=os.environ["ROUTERPLEX_API_KEY"],
    base_url="https://api.routerplex.com/v1",
)
 
response = client.chat.completions.create(
    model="glm-5.3-flash",
    messages=[{"role": "user", "content": "Extract the action items from these notes."}],
)
 
print(response.choices[0].message.content)

Claude Code can point at the same models with ANTHROPIC_BASE_URL=https://api.routerplex.com and model glm-5.3 or glm-5.3-flash. See the Claude Code setup guide.

Which GLM should you actually call? #

• Current GLM quality at the flagship rate — glm-5.3. Same dollars as 5.2, newer weights, 1M context.
• Volume, classification, or image input — glm-5.3-flash at $0.15 / $0.50. Do not budget the $0.075 promo.
• Already on 5.2 and evaluation shows no gain — stay. The sticker did not move.
• Need a cheaper 1M-context Flash from another lab — Qwen3.8 Flash at $0.15 / $0.47, text-only.

Give the key a hard budget. The live GLM-5.3 page is the number this route will actually bill.