ValueError / InvalidArgument
google.generativeai.types.generation_types.ModelNotFoundError OR ValueError: model not found
Stack trace
google.api_core.exceptions.NotFound: 404 Not Found Error code: 404 Error message: models/gemini-pro not found or is not available for use. Or: ValueError: model 'gemini-pro' is not a valid model identifier. Valid models: ['gemini-1.5-pro', 'gemini-2.0-flash', 'gemini-1.5-flash']
Why it happens
The google-generativeai SDK validates model names against Google's current API catalog when you call GenerativeModel(). Older model names like 'gemini-pro', 'gemini-pro-vision', or typos like 'gemini-2.0-pro' (which doesn't exist) are rejected. Additionally, if your API key doesn't have access to a particular model tier (e.g., Gemini 2.0 Flash requires paid API access), the model won't be found even if the name is valid.
Detection
Before calling GenerativeModel(model_name), validate the model name against the list returned by genai.list_models(). Add logging that prints which model you're requesting and which models are available in your API tier. Monitor API responses for 404 errors on model initialization.
Causes & fixes
Using deprecated model names: 'gemini-pro', 'gemini-pro-vision', or 'gemini-1.5-pro-latest'
Replace with current supported models: 'gemini-2.0-flash' (latest), 'gemini-1.5-pro', or 'gemini-1.5-flash'. Check available models with: genai.list_models() and look for models with 'gemini-' prefix.
Typo in model name: 'gemini-2.0-pro' (doesn't exist), 'gemini-1.5-pro-exp', or similar
Verify the exact model name using genai.list_models(). The correct current models are: 'gemini-2.0-flash', 'gemini-1.5-pro', 'gemini-1.5-flash'. Check spelling character-by-character, including hyphens.
API key lacks access to the requested model tier (e.g., Gemini 2.0 Flash on free tier)
Upgrade your Google Cloud project to a paid billing tier if requesting premium models. Free tier access is limited to 'gemini-1.5-flash' only. For Gemini 2.0 Flash, enable billing and verify the API key has the 'AI Platform' or 'Generative AI' permission.
Model name passed as None, empty string, or a variable that's undefined at runtime
Ensure the model name is a non-empty string set before GenerativeModel() is called. Use os.environ['GEMINI_MODEL'] with a default fallback: model_name = os.environ.get('GEMINI_MODEL', 'gemini-2.0-flash')
Code: broken vs fixed
import google.generativeai as genai
import os
genai.configure(api_key=os.environ['GOOGLE_API_KEY'])
# ❌ BROKEN: Using deprecated model name
model = genai.GenerativeModel('gemini-pro')
response = model.generate_content('Hello, world!')
print(response.text)import google.generativeai as genai
import os
genai.configure(api_key=os.environ['GOOGLE_API_KEY'])
# ✅ FIXED: Use current valid model name
model = genai.GenerativeModel('gemini-2.0-flash')
response = model.generate_content('Hello, world!')
print(response.text)
# OPTIONAL: Verify model exists first
available_models = [m.name for m in genai.list_models() if 'gemini' in m.name]
print(f'Available models: {available_models}')Workaround
If you must use a specific older model and can't upgrade, wrap GenerativeModel() in a try/except that catches the 404 error, then fall back to 'gemini-1.5-flash' as a default: try: model = genai.GenerativeModel(requested_model) except (ValueError, google.api_core.exceptions.NotFound): model = genai.GenerativeModel('gemini-1.5-flash') and log the fallback.
Prevention
Store the model name in an environment variable or config file, validate it against genai.list_models() at startup, and add integration tests that attempt to initialize the model before your app starts serving traffic. Use a config constant instead of hardcoded strings: GEMINI_MODEL = os.environ.get('GEMINI_MODEL', 'gemini-2.0-flash') and validate it once at boot time.