Gemma 3 from 270M to 27B, served from GPUs you control. The 27B model runs on a single 48GB card at FP8.
Pilot $4,900 fixed
$4,900 fixed. Pick the smallest size that passes your evaluation. The reference evaluation in the pilot is built to answer exactly that.
72 h$4,900 fixed
Gemma 3 live on your data in 72 hours. The clock starts after access and prerequisites.
/v1OpenAI-compatible
OpenAI-compatible endpoint. Your applications switch by changing the base URL.
Evalon your data
Reference evaluation on your data. Readout and production scoping.
We size the hardware and source it at cost, with no markup. Ballpark: one H100 80GB ≈ $25–30k; a 4×H100 node ≈ $180–240k; smaller models run on far less.
The model is one parameter. Everything else in your application stays as it is.
client = OpenAI(
base_url="https://llm.yourcompany.internal/v1", # inside your network
api_key="your-key",
)
reply = client.chat.completions.create(
model="gemma-3",
messages=[{"role": "user", "content": "Summarise this contract."}],
)Gemma 3 is released under Google's Gemma Terms of Use, not Apache 2.0. Commercial use is allowed, and Google's prohibited-use policy passes on to anything you build on it.
Gemma 4, released in March 2026, moved to Apache 2.0.
We read the current terms against your use when we scope the pilot.
Published by Google. We are not affiliated with the vendor. This is a summary, not legal advice.
Tell us the model, the data and the GPUs you have. An engineer from Limassol or Auckland scopes the 72-hour pilot with you.