Provenance record

Introducing MentalHealthBench

OpenAI (tier 1, lab) 2026-09-23T10:00:00.000Z Original ↗

Source note: Primary statements from a frontier lab. Self-interested by construction.

Discourse valence

Mixed or uncertain

confidence 70% · 1 item

-100

-50

+50

+100

Adverse readingFavourable reading

Single-model reading. This item did not meet the threshold for a multi-model panel.

AGI and timelines Evaluation and measurement Concentration of power

Excerpt as ingested

MentalHealthBench is an expert-informed benchmark for evaluating helpful and safe AI responses across realistic mental health conversations.

Every model that read this

Model	Provider	Stage	Score	Conf.	Latency	Prompt	When
Llama 3.3 70B	Meta	analysis	0	70%	4011ms	v1.0.0 / m1.0.1	2026-09-23 21:18

Llama 3.3 70B · reading

MentalHealthBench may improve AI safety and support.

evidence: speculative horizon: n/a

Evidence extracted

MentalHealthBench evaluates AI responses across mental health conversations

The chain

SOURCE     OpenAI (tier 1)
   ↓
DOCUMENT   c5cbbfa2-6015-482f-b401-63528310a5fe
           https://openai.com/index/introducing-mentalhealthbench
   ↓
EVIDENCE   1 extracted excerpt
   ↓
MODEL RUN  1 run, methodology 1.0.1
   ↓
SCORE      0  (Mixed or uncertain)
   ↓
CONFIDENCE 70%