Skip to main content

Supported Models

Kertex AI exposes a small set of tier aliases instead of raw Google model names. You call an alias such as kertexai-fast, and Kloudbean maps it to a managed Google Gemini model behind the scenes. This keeps your integration stable even when the underlying model is upgraded.

Kertex AI available models and pricing

Model tiers​

AliasTierBest for
kertexai-fastFastLow-latency chat and high-volume workloads.
kertexai-advancedAdvancedHighest reasoning quality for complex analysis.
kertexai-lightLightThe cheapest tier — classification, routing, short extraction tasks.
kertexai-embedEmbeddingsVector embeddings for retrieval / RAG (input billed only).

You can see the exact Google model each alias currently maps to, along with live pricing, in your console under the environment's Endpoint & SDK tab.

Pricing​

Pricing is per token, billed on what you actually use (input tokens + output tokens). Example published rates, shown per 1M tokens:

AliasInput (per 1M)Output (per 1M)
kertexai-fast$0.15$0.60
kertexai-advanced$1.25$5.00
kertexai-light$0.05$0.20
kertexai-embed$0.02$0.00
note

Rates shown here are examples. Your console's Endpoint & SDK tab always reflects the current pricing for your account. Embeddings bill on input only.

Choosing a tier​

  • Start with kertexai-fast for most chat and assistant workloads — it balances speed and cost.
  • Move specific calls to kertexai-advanced only where you need the strongest reasoning (complex analysis, hard multi-step tasks). It costs noticeably more per token.
  • Use kertexai-light for cheap, high-volume, low-complexity work such as classification, routing, tagging, or short summarisation.
  • Use kertexai-embed to turn text into vectors for semantic search and retrieval-augmented generation (RAG).

Managed model upgrades​

Because you call a Kloudbean alias rather than a pinned Google model id, Kloudbean can move a tier to a newer or better model without breaking your integration. Your code keeps sending kertexai-fast; the quality improves underneath. The current mapping is always visible in the Endpoint & SDK tab.

Next steps​