Completed evaluation
A result explicitly supplied by North ML from a completed run.
Benchmark transparency
Measured results.
Honest estimates.
North ML separates completed evaluation results from directional estimates. Precision without provenance is just decoration.
| Model | Evaluation | Score | Provenance |
|---|---|---|---|
| Aurora One Pro | General accuracy (25 questions)North ML internal evaluation | 79.20% | Measured |
| Aurora One Pro | Coding evaluationNorth ML internal evaluation | 42.80% | Measured |
| Aurora One Main | General capabilityDirectional family estimate | 55–68% | Estimated |
| Aurora One Main | Coding capabilityDirectional family estimate | 28–40% | Estimated |
| Aurora One Mini | General capabilityDirectional family estimate | 35–45% | Estimated |
| Forge-1 | Coding capabilityDirectional family estimate | 30–45% | Estimated |
Directional range
A planning estimate based on family positioning. Never shown as a precise tested result.
No claim
When trustworthy information is absent, the correct score is no score at all.