NOPE Evalsa fork of Weval

Open benchmarks for how AI handles people in distress

Transparent, reproducible, clinically-grounded tests of how public AI models behave in mental-health-adjacent and relational scenarios.

A NOPE public-good project · a fork of Weval

Leaderboard

Safety & Integrity

  • 1.
    C
    Claude 3.5 Haiku
    89%
  • 2.
    G
    Gemini 2.5 Flash
    89%
  • 3.
    O
    GPT 4o Mini
    78%

We measure safety and integrity by combining and averaging results across multiple evaluations, including those that measure hallucination, sycophancy, bias, and mental health crisis response.

View methodology

Leaderboard

Helpfulness & Reasoning

    We measure helpfulness and reasoning by combining and averaging results across multiple evaluations and dimensions: factual accuracy, helpfulness, coherence, depth, and argumentation.

    View methodology

    Evaluation

    Crisis response: passive suicidal ideation (multi-turn)

    Tests how a model responds to a user disclosing passive suicidal ideation over several turns. Grounded in C-SSRS principles: acknowledge without flattening, assess safety without interrogating, restrict means, and surface help without deflecting. Scenario is synthetic.

    mental-healthcrisissuicide
    3 models1 promptsJun 14, 2026

    NOPE Evals is a public-good project — a fork of the open-source Weval platform.