Models

A model that only does silicon

ChipCraftModel-1, or CCM-1: a capable model, custom-trained on ChipCraftX's verified silicon workflows and run on our own hardware. It builds RTL, tests it against real tools, and refines it from the results — privately, at a fraction of frontier-model cost.

Custom ModelEarly Access

ChipCraftModel-1

CCM-1

Our own model. Our own silicon data.

Describe it, and CCM-1 builds the RTL, tests it against real EDA tools, and uses the results to refine it — the whole job, not just the first draft. A capable model trained on verified silicon workflows nobody else has, and hosted on our own hardware so your unreleased design stays yours.

Everything it does
The whole job

Build. Test. Refine.

Generating a module was never the hard part. CCM-1 drafts to the interface you already have, builds in steps, gets it past real EDA tools, and repairs from the results — carrying the work past the first answer.

01Draft02Build03Pass04Wire
Pricing
$0.10
Input
per 1M tokens
$0.50
Output
per 1M tokens
5%
vs Sonnet 5
of the price
1%
vs frontier
of the price

Long context is never surcharged, and there is no premium tier for the same model.

Positioning

Beats Sonnet 5

A model that only has to be excellent at hardware does not pay for everything else. Indicative RTL positioning — not a published benchmark.

Private

Your design stays yours

Runs on ChipCraftX infrastructure. An unreleased architecture never goes to a third-party model provider, and is never used to train our models.

Data

Trained on verified workflows

Not just RTL — the whole workflow of building, simulating, and repairing it, filtered by real tools with benchmarks held out. A capable model that brings general reasoning to hardware, on data nobody else has.

Bring it your hardest module

Early access goes out to design teams first.

Request early access