GLM-5.3-Flash Uncensored API

Call GLM-5.3-Flash with Python

Install the OpenAI Python SDK, set your UNST33R key, and keep the model ID exactly as shown.

from openai import OpenAI

client = OpenAI(
    base_url="https://api.unst33r.si/v1",
    api_key="<your api key>",
)

response = client.chat.completions.create(
    model="unst33r-flash",
    messages=[{"role": "user", "content": "Explain this code."}],
)
print(response.choices[0].message.content)

Need the full setup? Follow the OpenAI Python base URL guide.

Call the API with curl

curl https://api.unst33r.si/v1/chat/completions \
  -H "Authorization: Bearer $UNST33R_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "unst33r-flash",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

Tool calling status

Tool calling is enabled for unst33r-flash. Send OpenAI-style function definitions in the tools array, execute the selected function in your application, then return its result to the model.

response = client.chat.completions.create(
    model="unst33r-flash",
    messages=[{"role": "user", "content": "Weather in Ljubljana?"}],
    tools=[{
        "type": "function",
        "function": {
            "name": "get_weather",
            "description": "Get current weather for a city",
            "parameters": {
                "type": "object",
                "properties": {"city": {"type": "string"}},
                "required": ["city"],
            },
        },
    }],
)

Self-hosted weights or a hosted uncensored API?

Abliterated and uncensored checkpoints of this family are also distributed as downloadable weights. Running them yourself means sourcing the weights, sizing a GPU box, serving an inference stack, and keeping it updated. A hosted API skips all of it:

• · No GPU rental or capacity planning — usage is billed per token.
• · No weights to download, store, or re-patch when a new revision ships.
• · OpenAI-compatible request shape — moving from a local server is mostly a base URL change.
• · Streaming and tool calling per the capability table above.

Self-hosting still wins when you need full weight custody or air-gapped inference. For everything else, unst33r-flash serves the same checkpoint behind an API key.