Environment details
- Programming language: Python
- OS: Linux (Kali Linux)
- Language runtime version: Python 3.13.12
- Package version: 2.0.1
Steps to reproduce
- List available models:
from google import genai
client = genai.Client(api_key="YOUR_API_KEY")
for model in client.models.list():
print(model.name)This returns Gemma models including:
models/gemma-4-26b-a4b-it
models/gemma-4-31b-it
- Attempt inference using the shorthand model alias:
from google import genai
client = genai.Client(api_key="YOUR_API_KEY")
response = client.models.generate_content(
model="gemma-4-31b-it",
contents="hello"
)
print(response.text)This returns:
ServerError
500 INTERNAL. {'error': {'code': 500, 'message': 'Internal error encountered.', 'status': 'INTERNAL'}}
- Attempt inference using the fully-qualified resource name:
from google import genai
client = genai.Client(api_key="YOUR_API_KEY")
response = client.models.generate_content(
model="models/gemma-4-31b-it",
contents="hello"
)
print(response.text)This succeeds correctly.
Additional observations:
- Gemini shorthand aliases work correctly. For example:
model="gemini-2.5-flash-lite"
works without requiring:
models/gemini-2.5-flash-lite
- The inconsistency appears specific to Gemma hosted inference endpoints.
- The returned 500 INTERNAL error is misleading because the issue appears related to model name resolution/routing rather than an actual server-side inference failure.
Expected behavior:
- Either shorthand aliases should work consistently for Gemma models,
- or the API/SDK should return a clear validation error indicating that fully-qualified model resource names are required.
Environment details
Steps to reproduce
This returns Gemma models including:
This returns:
This succeeds correctly.
Additional observations:
works without requiring:
Expected behavior: