Free Access to NVIDIA GLM‑4.7 and Minimax‑M2.1 with a Step‑by‑Step NIM Tutorial

What Is NVIDIA NIM?

NVIDIA Inference Microservices (NIM) is a cloud‑based inference platform that lets developers run large‑language models without installing any hardware. Although the official UI does not prominently list GLM‑4.7 and Minimax‑M2.1, the APIs for these models are already available as a hidden feature.

Step‑by‑Step Tutorial: Getting a Free API Key

1. Register and Log In

Visit the NIM integration homepage at https://build.nvidia.com/explore/discover. Create a new NVIDIA account with an email address if you do not already have one.

2. Phone Verification (Critical)

After registration, NVIDIA requires a phone‑number verification to prevent abuse. Selecting "China" and entering a +86 mobile number works perfectly; a verification code is sent and, once entered, unlocks free API access.

3. Obtain the API Key

Log in, open any model page (e.g., DeepSeek‑R1 or Llama‑3), and click the "Get API Key" button in the top‑right corner. The system generates a key prefixed with nvapi-. Save this key securely.

4. Call the Hidden Models

The UI does not expose a direct entry for GLM‑4.7, but you can invoke it by modifying the model parameter in the provided Python or curl example. Use one of the following identifiers:

GLM‑4.7: z-ai/glm4.7 Minimax‑M2.1: minimaxai/minimax-m2.1 Replace the original model name with one of the identifiers and run the request.

Example Request (Python‑style curl)

invoke_url='https://integrate.api.nvidia.com/v1/chat/completions'

authorization_header='Authorization: Bearer xxxxx'
accept_header='Accept: application/json'
content_type_header='Content-Type: application/json'

data='{
  "prompt": "",
  "messages": [{"role": "user", "content": "现有的鸡还是先有的蛋"}],
  "model": "z-ai/glm4.7",
  "temperature": 0.6,
  "top_p": 0.7,
  "frequency_penalty": 0,
  "presence_penalty": 0,
  "max_tokens": 4096,
  "stream": false
}'

response=$(curl --silent -i -w "
%{http_code}" \
  --request POST \
  --url "$invoke_url" \
  --header "$authorization_header" \
  --header "$accept_header" \
  --header "$content_type_header" \
  --data "$data")

echo "$response"

Running the script returns the model’s response along with the HTTP status code, confirming that the hidden GLM‑4.7 endpoint is reachable with the free key.