Vertex AI & Cloud GPU/TPU

Offering overview

Gemini is the headline, but Vertex AI on Google Cloud goes well beyond it. The platform's Model Garden exposes Gemini, Gemma, Imagen and Veo alongside partner models from Anthropic, Meta Llama, Mistral, OpenAI gpt-oss and others, with serverless per-token billing, batch pricing at half rate, provisioned throughput, tuning, evaluation, RAG Engine and an agent builder. Custom models can be trained and served on NVIDIA GPUs (L4, A100, H100, B200) or on Google's own TPU generations. Distinctive features are the TPU option for large-scale training, tight coupling with BigQuery and Vertex AI Vector Search, and controls such as data residency by region, CMEK and VPC Service Controls. Google Distributed Cloud extends the same model stack to on-premises and air-gapped sites. Data-heavy enterprises and teams building on Gemini are the natural fit. Pricing is usage based; Gemini 2.0 Flash-Lite on Vertex AI is listed at USD 0.075 per 1M input tokens and USD 0.30 per 1M output tokens, with cheaper batch rates.

Technical specifications

Containers

Unknown

Kubernetes

Unknown

Root access

Unknown

API access

Unknown

Autoscaling

Unknown

Persistent processes

Unknown

AI capabilities

Model serving

Gemini, Gemma, Imagen, Veo, Anthropic Claude, Llama, Mistral, OpenAI gpt-oss

AI frameworks

Not stated

Inference endpoints

Unknown

Fine-tuning

Unknown

Vector database

Unknown

Model catalog

Unknown

Monitoring

Not stated

Evaluation

Not stated

Security & compliance

Geography

Provider headquarters are separate from offering infrastructure and availability.

Available plans

Pricing can change. Promotional prices are shown separately from standard prices.