Edge
Run safety classification on your own infrastructure. No API calls, no conversation data leaving your environment — the same crisis-detection model that powers NOPE's Evaluate API, distributed as open weights.
Free for research, academic, and nonprofit use. Commercial licensing available.
How it reads
Edge emits structured XML: a <reflection> chain-of-thought, then zero or more <risk> tags.
It distinguishes genuine crisis from hyperbole and slang.
Input
"I want to end it all tonight"
Output
<reflection>User expresses intent…</reflection>
<risk subject="self" type="suicide" severity="high"/>
Input
"kms lmao missed that shot 💀"
Output
<reflection>Hyperbolic slang…</reflection><risks/>
Model variants
Same output format. Choose based on your accuracy and latency requirements.
nope-edge
RecommendedMaximum accuracy for production deployments.
- F1 score
- 90%
- Recall
- 86%
- Parameters
- 4B
- VRAM
- ~8GB
nope-edge-mini
SmallerFaster inference for high-volume environments.
- F1 score
- 83%
- Recall
- 75%
- Parameters
- 1.7B
- VRAM
- ~4GB
Benchmarked across NOPE's public test suites. See the full benchmarks →
What it detects
Nine risk types spanning crisis signals, interpersonal harm, and safeguarding concerns — each tagged with a subject (self / other / unknown), severity, and imminence.
suicide Ideation, planning, method-seeking, farewell behaviors
self_harm Non-suicidal self-injury, cutting, burning
self_neglect Disordered eating, medication non-adherence
violence Threats or intent to harm others
abuse Intimate partner violence, coercive control
sexual_violence Assault, harassment, coercion
exploitation Trafficking, sextortion, financial exploitation
stalking Unwanted pursuit, monitoring, harassment
neglect Child or elder neglect, failure to provide care
Edge is a detection aid, not a predictive, diagnostic, or therapeutic tool, and is not a replacement for clinical judgment.