NVIDIA NIM GLM-5.3 chat completions time out while /v1/models returns HTTP 200

AI & Data Science NVIDIA NIM Access/Accounts
1

Hello NVIDIA Developer Community,

I’m experiencing an issue with the NVIDIA NIM API when using the GLM-5.3 model.

Environment

  • OS: Ubuntu Linux
  • API endpoint: https://integrate.api.nvidia.com/v1
  • Model: z-ai/glm-5.3
  • Client: curl and Hermes Agent

Tests performed

  1. DNS resolution succeeds.
  2. TCP connectivity and TLS certificate verification succeed.
  3. GET /v1/models returns HTTP 200.
  4. POST /v1/chat/completions times out after 30 seconds with zero response bytes.
  5. Generating a new API key did not resolve the issue.
  6. Hermes Agent also hangs while waiting for model responses.

Could someone help me determine whether this is a GLM-5.3 inference-service issue, an account-access or entitlement issue, or another API problem?

I have not included my API key or other credentials. I can provide additional diagnostic output if needed.

Thank you.

2

This is happening to me but 3 minutes respond time for the deepseek v4.1 flash model…

3

Experiencing the same issue with deepseek-ai/deepseek-v4.1-flash and google/gemma-4-31b-it for a couple of days already. The requests time out after minutes of waiting. IP/TCP/DNS/HTTPS connectivity is in place. Authentication with API key works, JSON schemas are valid, but there is no output from the models.

Using the examples from the model description pages themselves (curl/shell):

curl --fail-with-body --show-error --no-buffer https://integrate.api.nvidia.com/v1/chat/completions 
-H "Content-Type: application/json" 
-H "Authorization: Bearer $NVIDIA_API_KEY" 
--data-binary @- <<JSON
{
"model": "deepseek-ai/deepseek-v4.1-flash",
"messages": [{"role":"user","content":[{"type":"text","text":"Which LLM model am I interacting with?"}] }],
"temperature": 1,
"top_p": 0.95,
"max_tokens": 262144,
"stream": false
}
JSON

curl: (22) The requested URL returned error: 504

stream=false
if [ "$stream" = true ]; then
accept_header='Accept: text/event-stream'
else
accept_header='Accept: application/json'
fi

cat > payload.json <<JSON
{"model":"google/gemma-4-31b-it","max_tokens":1024,"stream":false,"temperature":0.5,"top_p":1,"stop":null,"frequency_penalty":0,"presence_penalty":0,"seed":0,"messages":[{"role":"system","content":"You are a helpful assistant."},{"role":"user","content":"What is 17 * 23?"}]}
JSON

curl --fail-with-body --show-error --no-buffer  https://integrate.api.nvidia.com/v1/chat/completions 
-H "Authorization: Bearer $NVIDIA_API_KEY" 
-H "Content-Type: application/json" 
-H "$accept_header" 
-d @payload.json

curl: (22) The requested URL returned error: 504

Trying to access the models from Slovenia/EU, the famous SI country