Product comparison
Verdictan vs. Not Diamond
You need every request to reach an appropriate model without losing policy control or cost discipline.
Why Verdictan
Verdictan applies policy at the gateway and records request-level decision evidence for supported AI traffic. Run it in a self-hosted or air-gapped environment.
Key differences
Verdictan vs. Not Diamond Model Router: Key differences
Verdictan governs supported model traffic, while Not Diamond predicts which candidate model should answer each query.
Not Diamond returns a model recommendation
The application then calls the selected model through its preferred provider SDK.
Teams can tune quality, cost, and latency priorities
The pre-trained router supports named tradeoffs and a continuous cost-to-quality blend.
Custom routers learn from evaluation results
Teams provide representative inputs, candidate responses, and numerical scores from their evaluation process.
Side by side
Feature comparison
Scroll horizontally to review all comparison columns.
| Feature | Verdictan | Not Diamond Model Router |
|---|---|---|
| Routing boundary and integration | ||
| Model routing | Documented | Documented |
| Gateway-initiated upstream model calls | Documented | Not documented in the reviewed sources |
| Client interfaces | Documented | Documented |
| Routing evidence | Documented | Documented |
| Selection logic | ||
| Pre-trained routing | Documented | Documented |
| Custom routing | Documented | Documented |
| Tradeoff modes | Documented | Documented |
| Configurable cost-quality routing | Not documented in the reviewed sources | Documented |
| Models, data, and policy | ||
| Custom inference endpoints | Documented | Documented |
| Router training inputs | Documented | Documented |
| Dataset-based prompt optimization | Not documented in the reviewed sources | Documented |
| Runtime policy coverage | Documented | Not documented in the reviewed sources |
Common questions
What to ask before you decide
Does Not Diamond call the selected model?
The documented quickstart returns a model selection, then the application calls that model separately.
Can Not Diamond route to a custom endpoint?
Yes. Custom routers can include fine-tuned models, agent workflows, and arbitrary inference endpoints.
Can teams optimize for cost?
Yes. The pre-trained router supports cost mode and a continuous cost-to-quality blend.
What data trains a custom router?
Training uses representative prompts, candidate model responses, and numerical evaluation scores.
What should the pilot measure?
Measure selection quality, total inference cost, latency, provider errors, policy behavior, fallback behavior, and evidence completeness.
How we researched this page
We checked the official sources below on Aug. 11, 2026. Product scope, plan access, beta status and support can change.
- Confirm supported model versions, rate limits, router latency, training limits, data handling, and pricing.
- Evaluate the router on representative production distributions because domain quality determines useful cost savings.
- Test model selection and gateway policy independently because the products make different decisions.
- 1. Verdictan: Verdictan AI governance gateway
- 2. Verdictan: Verdictan product
- 3. Verdictan: Verdictan documentation overview
- 4. Verdictan: verdictan gateway run
- 5. Verdictan: OpenAI integration
- 6. Verdictan: Usage costs and budgets
- 7. Not Diamond: What is Not Diamond?
- 8. Not Diamond: Pre-trained router: Chat
- 9. Not Diamond: Key concepts
- 10. Not Diamond: Training a custom router
- 11. Not Diamond: Routing between custom models
- 12. Not Diamond: Supported models for routing
- 13. Not Diamond: Prompt optimization quickstart
Verdictan is not affiliated with Not Diamond. Product names can be trademarks of their respective owners.
Put the comparison to work
Try Verdictan with representative traffic after you compare the documented boundaries.
You select a routing and governance design that meets verified quality, cost, latency, policy, and evidence requirements.