# Vals Mental Health Bench (Mental Health Bench (Vals))

> Published chatbot failure rates across urgent safety, health advice, and relationship boundaries in simulated teen conversations.

Canonical page: https://benchlm.ai/benchmarks/valsmentalhealth

- Category: [external](/external)
- Last updated: Captured October 15, 2026

Evaluation results published by [Vals AI](https://www.vals.ai/benchmarks/mental-health). Published leaderboard aggregates and evaluation settings only. Dataset contents are not included.

## About Mental Health Bench (Vals)

- Year: 2026
- Tasks: Simulated teen conversations across three risk areas
- Format: Urgent safety failure rate
- Difficulty: Conversation safety evaluation
- Paper: [Mental Health Bench](https://www.vals.ai/benchmarks/mental-health)

The table shows the source's rounded urgent safety failure rate, with lower values indicating fewer failures in this evaluation. Health advice and relationship-boundary rates remain separate in the snapshot. These results are display only, do not measure clinical outcomes, and do not enter model rankings.

Mental Health Bench (Vals) is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (18 models)

| Rank | Model | Creator | Score |
|------|-------|---------|-------|
| 1 | [GPT-6.1 Sol](/models/gpt-6-1-sol) | OpenAI | 15% |
| 2 | [GPT-5.6 Luna](/models/gpt-5-6-luna) | OpenAI | 23% |
| 3 | [Claude Sonnet 5.5](/models/claude-sonnet-5-5) | Anthropic | 23% |
| 4 | [Muse Spark 1.3 (Extra High)](https://www.vals.ai/benchmarks/mental-health) | Unknown | 26% |
| 5 | [Qwen 3.8 Max](https://www.vals.ai/benchmarks/mental-health) | Alibaba | 26% |
| 6 | [Claude Opus 5.5](/models/claude-opus-5-5) | Anthropic | 27% |
| 7 | [DeepSeek V4.1 Flash 1017](/models/deepseek-v4-1-flash) | DeepSeek | 31% |
| 8 | [Mistral Large 4](/models/mistral-large-4) | Mistral | 33% |
| 9 | [Muse Spark 1.3 (Minimal)](https://www.vals.ai/benchmarks/mental-health) | Unknown | 33% |
| 10 | [MiMo-V2.6-Pro](/models/mimo-v2-6-pro) | Xiaomi | 40% |
| 11 | [Kimi K3](/models/kimi-k3) | Moonshot AI | 44% |
| 12 | [Grok 4.7](/models/grok-4-7) | xAI | 44% |
| 13 | [Grok 4.5](/models/grok-4-5) | xAI | 50% |
| 14 | [GLM-5.3](/models/glm-5-3) | Z.AI | 51% |
| 15 | [Gemini 3.6 Flash](/models/gemini-3-6-flash) | Google | 52% |
| 16 | [Gemini 4 Argon](/models/gemini-4-argon) | Google | 53% |
| 17 | [MiniMax M3](/models/minimax-m3) | MiniMax | 54% |
| 18 | [Gemini 3.8 Flash](/models/gemini-3-8-flash) | Google | 56% |

## FAQ

### What does Mental Health Bench (Vals) measure?

Published chatbot failure rates across urgent safety, health advice, and relationship boundaries in simulated teen conversations.

### Which model leads the published Mental Health Bench (Vals) snapshot?

GPT-6.1 Sol currently leads the published Mental Health Bench (Vals) snapshot with a score of 15%.

### How many models are evaluated on Mental Health Bench (Vals)?

The Captured October 15, 2026 contains 18 AI models.

### Does Mental Health Bench (Vals) affect BenchLM's overall score?

Not directly. Mental Health Bench (Vals) is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.