Vertex AI Pricing Plans & Tiers

Vertex AI's pricing documentation lists 12 tiers with cryptic names ('Gemini 2.0 Flash, $10 per 1000 searches, $14 per 1000 search queries, $14 per 1000 queries, $25 per 1000 grounded prompts') but provides zero context for differentiation. This is not consumer-friendly pricing presentation but rather a technical reference that assumes prior knowledge of Gemini's inference variants and grounding mechanisms. The structure reflects Google Cloud's infrastructure-first pricing philosophy: different pricing tiers correspond to different model serving strategies (standard, cached, grounded) rather than feature tiers or usage buckets.

The granularity of published pricing ($2-$45 per 1K operations/tokens/queries) provides individual unit costs but masks total cost complexity—organizations cannot estimate monthly bills without understanding their workload composition (input tokens vs. grounded queries vs. cached tokens). This is fundamentally opaque pricing designed for enterprise customers with dedicated GCP account managers who can model costs as part of pre-sales engineering.

Absence of a Free tier or starter credits (unlike OpenAI, Anthropic, and Google's own Gemini API) creates a Vertex AI adoption barrier. Teams evaluating Gemini on Vertex must commit to GCP cost ($0.10+ minimum) before running any workload. This is intentional: Vertex is positioned as an enterprise ML platform, not a developer API. Individual developers use Claude, Gemini web, or OpenAI API; enterprise teams use Vertex when they're already embedded in Google Cloud.