Provider evaluation / Safety & wellbeing · Child safety
The October GPT-6 Sol and Luna system card reports improvements in some mental-health measures alongside regressions on some standard safety tests. It describes dedicated under-18 tests and dynamic simulated conversations for emotional reliance, mental health and self-harm.
Why it matters
Useful for distinguishing model evaluations from the behaviour of a deployed service and its additional safeguards.
Read with care
Provider-run evaluations and interpretation, not independent validation or clinical outcomes. Some tests omit system-level interventions. Scores from other system cards may not be comparable because measurement changes. This new card does not resolve the separate Common Sense Media assessment of earlier product versions. Selected narrative sections inspected; numerical charts and raw evaluation data not independently audited.
What was checked
Source record and stated publication date checked; see reading cautions for the extent and limits of the review. Checked 2026-10-12.