GLM API access — OpenAI-compatible, one key

DaoXE fronts Zhipu's GLM family through the standard OpenAI shape: POST /v1/chat/completions with a GLM model ID, or /v1/responses where the ID supports it. Existing OpenAI SDK code works unchanged — only the base URL and key change. How the cost split works against official APIs: cheapest APIs guide.

Call GLM through DaoXE#

Chat path: /v1/chat/completions with Authorization: Bearer. Reasoning-capable GLM IDs spend tokens on thinking before the answer — give max_tokens enough headroom or the reply comes back empty with finish_reason: "length".

curl --fail-with-body --show-error --silent \
  https://daoxe.com/v1/chat/completions \
  -H "Authorization: Bearer ${DAOXE_API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "YOUR_EXACT_MODEL_ID",
    "max_tokens": 64,
    "messages": [{"role": "user", "content": "Say hello in one sentence."}]
  }'

# Reasoning-capable GLM IDs spend tokens on thinking BEFORE the answer.
# A small max_tokens returns empty content with finish_reason="length" -
# budget for the thinking, not just the answer.
from openai import OpenAI

client = OpenAI(base_url="https://daoxe.com/v1", api_key="YOUR_DAOXE_KEY")
r = client.chat.completions.create(
    model="YOUR_EXACT_MODEL_ID",   # a glm-* id from GET /v1/models
    max_tokens=2048,               # room for thinking + answer
    messages=[{"role": "user", "content": "2+2? Think briefly, then answer."}],
)
msg = r.choices[0].message
print(getattr(msg, "reasoning_content", None))  # thinking, if exposed
print(msg.content)                               # final answer
print(r.choices[0].finish_reason)                # "stop" means it fit

How GLM exposes reasoning#

Reasoning-capable GLM IDs treat max_tokens as the budget for thinking and answer: the chain of thought runs inside the completion, tokens counted against the same cap. Undersize the cap and you get empty content with finish_reason: "length" — no error, just a blank. Size for the thinking first. Some IDs expose the trace via reasoning_content on the message; treat it as best-effort and read the catalogue notes rather than assuming.

GLM-specific notes#

• Reasoning models eat budget. Thinking tokens count toward max_tokens; a small cap yields an empty content body, not an error. Size the budget for the thinking, not the answer.
• No native Zhipu endpoint shape. There is no ep- style ID here — use catalogue IDs from GET /v1/models.
• Exact IDs from the catalogue. GLM versions move fast; copy the ID from your account's live list rather than a blog post.

What it costs#

Verify you actually get the model#

Prove the endpoint works before blaming the client — if this fails, no setting will fix it:

export DAOXE_API_KEY="your_api_key"

# List the exact model IDs your account can call
curl --fail-with-body --show-error --silent \
  https://daoxe.com/v1/models \
  -H "Authorization: Bearer ${DAOXE_API_KEY}"

Confirm connectivity, then diff a fixed prompt at temperature 0 against the official Zhipu API to make sure the tier matches:

Frequently asked questions#