GLM 5.3, for long coding and agent work
Z.ai's GLM-5.3 keeps GLM-5.2's base model and gets its gains from post-training, aimed at complex coding and long-horizon tasks. This endpoint serves it text in, text out, with reasoning on every request.
- 743B MoE
- 256K context
- Tool calling
- Open weights
Per token, nothing else.
Input / output per 1M tokens
$0.079 / $3.00
openrelay/glm-5.3Full catalog on the inference pricing page. Add a card and make a deposit to start.
from openai import OpenAI
client = OpenAI(
base_url="https://inference.openrelay.inc/v1",
api_key="or_••••••••", # same SDK, new base URL
)
resp = client.chat.completions.create(
model="openrelay/glm-5.3",
messages=[{"role": "user", "content":
"Find the bug: def last(xs): return xs[len(xs)]"}],
max_tokens=8192, # room for the reasoning and the answer
)
print(resp.choices[0].message.content)The spec sheet.
What the model is, what it takes in, and the surface it serves on.
- Model id
- openrelay/glm-5.3
- Parameters
- 743B total, about 40B active per token (mixture-of-experts)
- Architecture
- DeepSeek Sparse Attention, 256 routed experts with 8 active per token plus one shared
- Modality
- Text in, text out
- Context
- 256K tokens (262,144) on this endpoint, prompt and completion combined. The weights support 1M.
- Endpoint
- POST /v1/chat/completions with tool calling and reasoning
- Precision
- MXFP4
- License
- GLM-5.3 License, open weights: MIT-style terms, plus a Z.ai security review for a model-as-a-service operator whose revenue with its affiliates passes 10 billion US dollars in any 12 months
Where GLM 5.3 earns its place.
What this model is actually for, versus the rest of the catalog.
Post-training on a known base
GLM 5.3 runs on the same base model as GLM 5.2. Z.ai credits every gain to post-training, aimed at complex coding and long-horizon tasks.
Reasoning on every request
The model thinks before it answers on every call, and that reasoning bills as output. Leave room for it in max_tokens.
Next to GLM 5.2 and 5.3 Flash
GLM 5.2 serves a 1M-token window and the Anthropic Messages API. GLM 5.3 Flash takes images at 18B active. Switching between them is a model id.
GLM 5.3, answered.
Is there a GLM 5.3 API?
Yes: openrelay/glm-5.3, on the OpenAI-compatible chat completions endpoint at inference.openrelay.inc/v1, with tool calling. The OpenAI SDK works with a new base URL and an OpenRelay key.
What does the GLM 5.3 API cost?
The rate on this page is read from the catalog when the page renders. Input and output tokens bill separately, a cached prompt prefix bills at a lower input rate, and reasoning counts as output.
GLM 5.3, GLM 5.3 Flash or GLM 5.2?
GLM 5.3 for complex coding and long-horizon agent work. GLM 5.3 Flash for image input. GLM 5.2 for the 1M-token window or the Anthropic Messages API. GLM 5.3 speaks chat completions only.
What license are the weights under?
The GLM-5.3 License: MIT-style permission to use, modify and serve the weights, plus one condition. A company that runs a model-as-a-service business, and whose revenue together with its affiliates passes 10 billion US dollars in any 12 months, must pass Z.ai's security review before commercial use.
First request in five lines.
Point the OpenAI SDK at inference.openrelay.inc/v1 and run GLM 5.3 per token. No contract, no minimums.