BigModel (Zhipu AI)

Available from v1.109.0.

Use bigmodel/ for the domestic BigModel platform and zai/ for international Z.AI. Each platform uses its own credentials and default URL

API Key

Create an API key in your BigModel account and set it in the environment where you run LiteLLM

export BIGMODEL_API_KEY="YOUR_BIGMODEL_API_KEY"
pip install litellm

Sample Usage

from litellm import completion

response = completion(
    model="bigmodel/glm-5.3",
    messages=[{"role": "user", "content": "Hello"}],
    thinking={"type": "enabled"},
    reasoning_effort="low",
)
print(response.choices[0].message.content)
print(response.usage)

You do not need to set api_base. A successful call returns an assistant reply and token usage

Sample Usage - Streaming

from litellm import completion

stream = completion(
    model="bigmodel/glm-5.3",
    messages=[{"role": "user", "content": "Hello"}],
    reasoning_effort="low",
    stream=True,
    stream_options={"include_usage": True},
)
for chunk in stream:
    if chunk.choices:
        print(chunk.choices[0].delta.content or "", end="")
    if getattr(chunk, "usage", None):
        print(chunk.usage)

Responses API

from litellm import responses

response = responses(
    model="bigmodel/glm-5.3",
    input="Hello",
    reasoning={"effort": "low"},
    store=True,
)
print(response.output)
print(response.usage)

Set store=True if you need continuation, retrieval, or deletion, and use the response ID returned by LiteLLM. BigModel documents a seven-day retention period and does not support cancel, compact, native WebSocket, or file_search

Anthropic Messages API

Use the asynchronous Messages SDK entry point

import asyncio
import litellm

async def main() -> None:
    response = await litellm.anthropic.messages.acreate(
        model="bigmodel/glm-5.3",
        max_tokens=1024,
        messages=[{"role": "user", "content": "Hello"}],
        thinking={"type": "adaptive"},
        output_config={"effort": "low"},
    )
    print(response.content)
    print(response.usage)

asyncio.run(main())

Each API retains its native response format and streaming events. Messages preserves native thinking, output_config, and reasoning_effort parameters

Models and pricing

Prefix the model ID with bigmodel/. These three catalog models support all three APIs and use always-on thinking. GLM-5.3 accepts low, high, and max reasoning effort

The following prices are CNY per million tokens, checked against official pricing on 2026-10-27

Model ID	Input	Cached input	Output
glm-5.3	8	2	28
glm-5.3-flash	0.8	0.23	2.8
glm-5.3-flashx	2	0.57	7

LiteLLM stores USD/token prices using a static 7.0 CNY/USD conversion, with the source and verification date in the catalog metadata. Use existing litellm_params cost fields for deployment-specific contract prices

Using with LiteLLM Proxy

Install the proxy with pip install 'litellm[proxy]'. Set BIGMODEL_API_KEY and a strong LITELLM_MASTER_KEY in the gateway environment, then save this as config.yaml

model_list:
  - model_name: glm
    litellm_params:
      model: bigmodel/glm-5.3
      api_key: os.environ/BIGMODEL_API_KEY

general_settings:
  master_key: os.environ/LITELLM_MASTER_KEY

litellm --config config.yaml --host 127.0.0.1 --port 4000

In a second terminal with the gateway key set, send a request

curl http://localhost:4000/v1/chat/completions \
  -H "Authorization: Bearer $LITELLM_MASTER_KEY" \
  -H 'Content-Type: application/json' \
  -d '{"model":"glm","messages":[{"role":"user","content":"Hello"}],"reasoning_effort":"low"}'

curl http://localhost:4000/v1/responses \
  -H "Authorization: Bearer $LITELLM_MASTER_KEY" \
  -H 'Content-Type: application/json' \
  -d '{"model":"glm","input":"Hello","reasoning":{"effort":"low"}}'

curl http://localhost:4000/v1/messages \
  -H "Authorization: Bearer $LITELLM_MASTER_KEY" \
  -H 'Content-Type: application/json' \
  -d '{"model":"glm","messages":[{"role":"user","content":"Hello"}],"max_tokens":1024,"thinking":{"type":"adaptive"},"output_config":{"effort":"low"}}'

The same deployment name serves all three routes. Each response contains the reply and native token usage. Keep the BigModel key on the gateway; clients use the LiteLLM key

Codex works through /v1/responses with no extra settings, since LiteLLM forwards the include and parallel_tool_calls fields it sends

API Base

api_base is optional and defaults to https://open.bigmodel.cn. Set BIGMODEL_API_BASE to override the default, or pass an explicit api_base to override the environment value. API-specific bases and complete endpoint URLs are normalized, preserving a custom proxy's path prefix

A GLM Coding Plan key uses https://open.bigmodel.cn/api/coding/paas/v4 as its api_base. Chat Completions then calls the Coding Plan endpoint, and Responses and Messages keep the endpoints in the table below

Proxy endpoint	BigModel endpoint
/v1/chat/completions	/api/paas/v4/chat/completions
/v1/responses	/api/v1/responses
/v1/messages	/api/anthropic/v1/messages

In the Admin UI, select BigModel (Zhipu AI). Its credential form requires an API Key and provides the default API Base

See the official Chat Completions, Responses, and Messages guides for provider API details