Run Nano Banana 2 and Nano Banana Pro via one API — Puzzle
One endpoint. A fraction
of the cost.
The API
Cost
Problem
Teams pay full retail API rates on every request to Google Cloud and Vertex AI. Costs scale linearly as you grow — with no discount in sight.
Self-hosting an inference layer means huge upfront cost, a DevOps team to keep it alive, and no real access to Google's models.
Overpay at retail, or run inference you never wanted to own. Puzzle sits in between: the same Google models, billed on usage.
One
Endpoint
One POST to Puzzle. Keep your prompts. Google runs the live models — you don't operate the cluster.
Our
Network
Volume rates from Google, billed to you on usage. Fund with USDC/USDT or a USD wire; credit lands after confirmation. Limits are per account.
Why
Puzzle
Wins
GOOGLE
MODELS,
ONE API
NANO BANANA 2 / PRO

VEO

OMNI

GEMINI

Frequently
Asked
Questions
Nano Banana 2 and Nano Banana Pro are live. Omni, Veo, and Gemini return 422 until they ship. Status is in the console and docs.
Sign up, create a key, POST /v1 with a Bearer token. Curl is in the docs. Keep your prompts — pointing an official Google SDK at our host will not work.
Crypto is self-serve in the console: USDC or USDT on Ethereum, Tron, or BNB. Log the hash after you send; we credit GET /v1/balance when we confirm the received amount. Prefer a bank transfer, not crypto? WhatsApp or Telegram @PUZZLE_Support_AI — we invoice and credit the same balance. No card checkout.
The key is valid and confirmed credit is $0.00. Fund, wait for confirmation, then retry. 402 is expected — not a down endpoint.
No. Puzzle does not use your prompts or outputs to train Google models or ours.
