Moved out client configuration to Vertex AI for Gemini 2.5 Flash. Comparing calls between old configuration and new, we are seeing results to be wildly different and prompt size to be almost double. Config is the same.
Inspecting prompt on BAML Studio, both prompts look the same so not sure where the size difference is coming from.
Anyone seen a similar behavior?
Config:
client
GeminiFlash25Vertex {
provider vertex-ai
options {
model gemini-2.5-flash
location us-west1
project_id env.VERTEX_PROJECT_ID
credentials env.GOOGLE_APPLICATION_CREDENTIALS
generationConfig {
temperature 1.0
thinkingConfig {
includeThoughts false
thinkingBudget 0
}
}
}
}