Models

Three models. All private. All on our hardware.

Every model runs on dedicated hardware we own in the United Kingdom. Zero retention on all models: not stored, not logged, not trained on.

Zero retention on all models. Your API request is processed in memory and discarded. No model ever trains on your data.

Hush Omega

Everyday balance

Capability and speed in balance. Built for everyday professional work: drafting, summarising, correspondence, general analysis.

model: "omega"

Best for

Email drafting, document summaries, clinical correspondence, client communication, general questions.

Hush Prism

Fastest

Optimised for speed. Near-instant responses even at high volume. Ideal for real-time applications, chat interfaces, and high-throughput pipelines.

model: "prism"

Best for

Chat applications, real-time suggestions, form auto-fill, classification, high-volume processing.

Hush Lambda

Deep reasoning

Engineered for complex, multi-step reasoning and rigorous analysis. Takes longer but produces more thorough, nuanced results.

model: "lambda"

Best for

Contract analysis, regulatory research, code generation, multi-step reasoning, complex document review.

At a Glance

Model comparison

Model	API ID	Strength	Speed	Input price	Output price
Hush Omega	`omega`	Balanced capability	Fast	£0.50 / 1M	£1.50 / 1M
Hush Prism	`prism`	Speed optimised	Fastest	£0.50 / 1M	£1.50 / 1M
Hush Lambda	`lambda`	Deep reasoning	Thorough	£0.50 / 1M	£1.50 / 1M

All prices in GBP. All models share the same pricing. Full pricing details and calculator →

Choosing a Model

Which model should I use?

Start with Omega. It handles the vast majority of professional workloads well: drafting, summarising, Q&A, general analysis. Most applications should default to Omega.

Switch to Prism when speed is critical: real-time chat, auto-complete, classification at scale, or any pipeline where latency matters more than depth.

Switch to Lambda for hard problems: multi-step legal reasoning, complex contract analysis, code generation, research synthesis. It takes longer but the reasoning is deeper.

All models are available on all plans. You can switch between models per request. Just change the model parameter.

Privacy

What happens to your data

The same thing happens on every model: nothing.

Your API request arrives at our hardware. It is processed in memory. The response is returned. The request data is discarded. There is no log, no copy, no training set. This is not a policy that can be changed by a future terms update; it is a consequence of the architecture. There is no storage layer for API requests to be written to.

This applies uniformly to Omega, Prism, and Lambda. There is no model tier where retention differs.

Verify this yourself →

Start building

Create a free account and try every model. No card required.

Create a free account and copy your key →

API reference · Pricing · Quickstart · Changelog