GLM 5.3 on OpenRouter: Model ID, Provider Prices & Setup

GLM 5.3 is live on OpenRouter as z-ai/glm-5.3, with 47 endpoints from $0.07 to $2.80 per 1M input tokens. Compare providers, then copy a working curl or Python request.

Quick answer: Yes, GLM 5.3 is on OpenRouter. The model ID is z-ai/glm-5.3, and it has a context window of about 1M tokens. When we checked on November 8, 2026, OpenRouter listed 47 endpoints for it, with input prices from $0.07 to $2.80 per 1M tokens. Z.AI's own endpoint costs $1.40 input / $4.40 output. Because endpoints differ on price and maximum output, pin a provider for anything that matters.

This article first appeared on launch day, when GLM 5.3 was not yet listed. OpenRouter's catalog dates the listing to August 18, 2026, and the guide below now reflects what is available today.


GLM 5.3 on OpenRouter at a Glance

QuestionAnswer
Model IDz-ai/glm-5.3
Other IDsz-ai/glm-5.3:batch, z-ai/glm-5.3-prime, and the ~z-ai/glm-latest alias
InputText only
Context window1,048,576 tokens on most endpoints
Max output128,000 to 943,718 tokens, depending on the endpoint
Z.AI first-party price$1.40 input / $4.40 output per 1M tokens
ThinkingAlways on; it cannot be disabled

Prices on OpenRouter move often, so treat the numbers here as a snapshot and check the live endpoints API before committing to a budget.

Provider Prices

This is a selection of the 47 endpoints, taken from OpenRouter's endpoints API on November 8, 2026. All prices are per 1M tokens.

Endpoint tagInputOutputMax output
wafer/us$0.07$1.40943,718
inference-net/us$0.08$3.60943,718
reka$0.14$2.10262,144
deepinfra/fp4$0.56$2.50131,072
novita/fp8$0.63$1.98131,072
cloudflare$0.70$2.20943,718
siliconflow/fp8$0.70$2.20262,144
z-ai/fp8 (first party)$1.40$4.40131,072
fireworks$1.40$4.40943,718
amazon-bedrock/us-east-1$1.68$5.28131,072
alibaba/prime$2.80$8.80131,072

Three things stand out in the full list:

  • The input price is not the whole bill. wafer/us and relace/fp4 have almost the same input price, but relace/fp4 charges about three times as much for output ($4.38 against $1.40). GLM 5.3 always reasons, so output tokens often drive the cost.
  • Max output varies by a factor of seven. A long agent run that fits on a 943,718-token endpoint can be cut off on a 128,000-token one. Z.AI's documentation states 128K as the model's maximum output, so don't assume a higher figure in the catalog applies to your workload without testing it.
  • The tag can signal a different deployment. Many of the cheapest endpoints carry tags such as fp4, which usually indicates a lower-precision quantized deployment. Run your own evaluation before moving production traffic to one.

For a wider cost comparison, including the Coding Plan, see our GLM 5.3 pricing guide.

How to Call GLM 5.3 Through OpenRouter

OpenRouter uses the standard OpenAI-compatible endpoint, so a basic request looks like this:

curl https://openrouter.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "z-ai/glm-5.3",
    "messages": [{"role": "user", "content": "Find the race condition in this function."}]
  }'

Pin a provider

Without a provider block, OpenRouter chooses the endpoint for you. To control it, name the providers you have tested and turn off fallbacks:

curl https://openrouter.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "z-ai/glm-5.3",
    "provider": { "order": ["Z.AI"], "allow_fallbacks": false },
    "messages": [{"role": "user", "content": "Refactor this module and explain the changes."}]
  }'

With allow_fallbacks set to false, you get the provider you asked for or an error. That is usually what you want for long-context work, where a silent switch to a smaller output limit is worse than a clean failure.

Python

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://openrouter.ai/api/v1",
    api_key=os.environ["OPENROUTER_API_KEY"],
)

response = client.chat.completions.create(
    model="z-ai/glm-5.3",
    messages=[{"role": "user", "content": "Write tests for this parser."}],
    extra_body={"provider": {"order": ["Z.AI"], "allow_fallbacks": False}},
)

print(response.choices[0].message.content)

Moving From GLM 5.2

Changing the model ID is most of the work, with one exception: GLM 5.3 always runs with reasoning on. Z.AI's documentation says a request with thinking.type: "disabled", which GLM 5.2 accepts, will fail on 5.3, so remove that setting before you switch. Z.AI offers three effort levels: low, high, and max.

Our GLM 5.3 migration guide covers the remaining differences. If you are still on the older model, the GLM 5.2 OpenRouter guide covers its setup.

OpenRouter or Z.AI Direct?

Choose OpenRouter if you want one API key across many models, automatic failover, or access to cheaper third-party endpoints.

Choose the Z.AI API directly if you want the first-party deployment and Z.AI's own reasoning parameters without a routing layer in between.

If you need a faster, cheaper model and multimodal input, GLM 5.3 Flash is a separate listing. See GLM 5.3 Flash on OpenRouter.

FAQ

Is GLM 5.3 available on OpenRouter? Yes. It is listed as z-ai/glm-5.3, and OpenRouter's catalog dates the listing to August 18, 2026.

How much does GLM 5.3 cost on OpenRouter? It depends on the endpoint. On November 8, 2026, input ranged from $0.07 to $2.80 per 1M tokens and output from $1.40 to $8.80. Z.AI's own endpoint was $1.40 input and $4.40 output.

Is GLM 5.3 free on OpenRouter? No. There was no :free variant of GLM 5.3 in the catalog on November 8, 2026.

What is the difference between z-ai/glm-5.3 and z-ai/glm-5.3-prime? They are separate catalog entries. z-ai/glm-5.3-prime was listed at $2.80 input and $8.80 output, double Z.AI's standard rate. Check its model page for current details before you choose it.

Does GLM 5.3 support tool calling on OpenRouter? Yes on the endpoints we checked, including the cheapest twelve and z-ai/fp8. Support is reported per endpoint, so look for tools in supported_parameters in the endpoints API.

Can I turn off thinking? No. Z.AI's documentation says GLM 5.3 only supports reasoning enabled, so plan for reasoning tokens in your output budget.


Sources

Last updated: November 8, 2026

Start Using GLM 5 Today

Try GLM 5 free — reasoning, coding, agents, and image generation in one platform.