Not Diamond Code
Not Diamond builds intelligent model-routing infrastructure. Not Diamond Code is our model router purpose-built for long-horizon coding-agent workloads. It automatically selects the model and reasoning effort for each step of a session, helping teams maintain frontier-level quality at 20%+ lower inference cost.
Early access
Not Diamond Code is currently available through a gated waitlist.
Request access here. Our team will follow up to enable your workspace.
Quick Links
Section titled “Quick Links”- Getting Started: Install, set up, and run your first session
- Data & Privacy: Understand our data handling and compliance
How Not Diamond Code works
Section titled “How Not Diamond Code works”At each step of a coding-agent session, Not Diamond Code evaluates the broader session, including expected downstream work and cache state, to select the model and reasoning effort that optimize expected quality and total session cost. It uses more capable models when they are likely to improve the outcome or finish in fewer steps, and more efficient models when additional intelligence is unlikely to add value.
A lightweight local proxy runs alongside the coding-agent harness and computes the derived metadata needed for routing. Only that metadata is sent to the Not Diamond optimization service, which returns a routing recommendation. The model request itself runs through your configured provider or gateway.
Raw prompts, code, inputs, and outputs are not sent to Not Diamond. Recommendations continuously adapt based on developer feedback signals.
To get started with Not Diamond Code, see Introduction. For details on the metadata sent to Not Diamond and how it is retained, see Data Dictionary and Data Retention.
Need help?
Section titled “Need help?”See Troubleshooting or contact our team support@notdiamond.ai.