Congrats to the team to provide the new Gemma 3 models and the new endpoint on the Google AI API. The announcement blog - https://blog.google/technology/developers/gemma-3/ - reads wonderfully. Until someone puts it to the test...
Create AI with advanced text and visual reasoning capabilities
Easily build applications that analyze images, text, and short videos, opening up new possibilities for interactive and intelligent applications
HTTP 400: "Image input modality is not enabled for models/gemma-3-27b-it"
HTTP 400: "Audio input modality is not enabled for models/gemma-3-27b-it"
Tried different images (png, jpg, bmp), video (mp4) and PDF documents - both via inlineData and per File API using fileData attributes.
Create AI-driven workflows using function calling
Gemma 3 supports function calling and structured output to help you automate tasks and build agentic experiences.
HTTP 400: "Function calling is not enabled for models/gemma-3-27b-it"
HTTP 400: "Json mode is not enabled for models/gemma-3-27b-it"
HTTP 400: "Enum mode is not enabled for models/gemma-3-27b-it"
System Instruction
HTTP 400: "Developer instruction is not enabled for models/gemma-3-27b-it"
Code execution?
It's not explicitly mentioned in the blog. What's the situation here? Right now...
HTTP 400: "Code execution is not enabled for models/gemma-3-27b-it"
Candidate count > 1?
It's not explicitly mentioned in the blog. What's the situation here? Right now...
HTTP: "Multiple candidates is not enabled for models/gemma-3-27b-it"
Not sure what's the issue is...
However, with the announcement blog I would kind of expect that the mentioned features are available and operational from day 0 on. And not hoping for the best and then those are being added at a later stage.
I'm not sure whether those features have been disabled for the gemma3 model in the Gemini API and the model itself might be capable to deal with everything, like when used locally or deployed in Vertex AI or Cloud Run with GPU...
What are your observations?
