DistressBench leaderboard
As of 16 Nov 2026, the top score on DistressBench is 88.4% by Muse Spark 1.2 (Meta), ahead of Muse Spark Open at 87.7%, across 43 model configurations with a published result (Scale AI SEAL, result dated 28 Oct 2026). Quality of crisis support in 718 clinician-written conversations about suicide, self-harm and acute distress, scored against clinicians' rubrics. Scale AI. DistressBench is shown for reference only and does not count toward the BenchLeader Index.
Key facts (as of 16 Nov 2026)
- Muse Spark 1.2 holds the top DistressBench score at 88.4% (Scale AI SEAL, 28 Oct 2026).
- Muse Spark Open is second on DistressBench at 87.7% (Scale AI SEAL, 28 Oct 2026).
- DistressBench: Quality of crisis support in 718 clinician-written conversations about suicide, self-harm and acute distress, scored against clinicians' rubrics. Scale AI (Scale AI SEAL).
- 43 model configurations have a published DistressBench result on BenchLeader (BenchLeader, 16 Nov 2026).
- DistressBench does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).
Quality of crisis support in 718 clinician-written conversations about suicide, self-harm and acute distress, scored against clinicians' rubrics. Scale AI.
- Published by
- Scale AI SEAL
- Category
- Safety & honesty
- Index weight
- Reference only
- Models
- 43
- Data as of
- 16 Nov 2026
- Newest result
- 12 Nov 2026
Scale AI SEAL Leaderboards.
- 1Muse Spark 1.288.4%
- 2Muse Spark Open87.7%
- 3Muse Wonder85.9%
- 4Muse Spark 1.384.9%
- 5Inkling 2 Small82.3%
- 6DeepSeek V4.1 Flash79.8%
- 7Inkling78.5%
- 8Qwen3.8 Max (max)77.6%
- 9Qwen3.8 2.4T A95B77.5%
- 10Mistral Large 477.0%
- 11GPT-5.6 Terra76.4%
- 12Claude Opus 5.575.6%
- 13Claude Sonnet 5.574.6%
- 14GPT-6 Astra74.3%
- 15GPT-6.1 Sol74.2%
| # | ||||
|---|---|---|---|---|
| 1 | 88.4% | – | 2026-10-28 | |
| 2 | 87.7% | – | 2026-10-28 | |
| 3 | 85.9% | 64.4 | 2026-10-28 | |
| 4 | 84.9% | 61.6 | 2026-10-28 | |
| 5 | 82.3% | 57.9 | 2026-11-12 | |
| 6 | 79.8% | – | 2026-10-28 | |
| 7 | 78.5% | 50.4 | 2026-10-28 | |
| 8 | 77.6% | 58.7 | 2026-10-28 | |
| 9 | 77.5% | 51.4 | 2026-10-28 | |
| 10 | 77.0% | 52.9 | 2026-10-28 | |
| 11 | 76.4% | – | 2026-10-28 | |
| 12 | 75.6% | 61.0 | 2026-10-28 | |
| 13 | 74.6% | – | 2026-10-28 | |
| 14 | 74.3% | 66.4 | 2026-10-28 | |
| 15 | 74.2% | 62.0 | 2026-10-28 | |
| 16 | 73.6% | 54.3 | 2026-10-28 | |
| 17 | 73.5% | 60.7 | 2026-10-28 | |
| 18 | 72.7% | 47.2 | 2026-10-28 | |
| 19 | 72.7% | 57.8 | 2026-10-28 | |
| 20 | 72.5% | – | 2026-10-28 | |
| 21 | 71.7% | – | 2026-10-28 | |
| 22 | 71.5% | 66.4 | 2026-10-28 | |
| 23 | 71.4% | 58.7 | 2026-10-28 | |
| 24 | 71.2% | 54.9 | 2026-10-28 | |
| 25 | 70.7% | 56.4 | 2026-10-28 | |
| 26 | 70.2% | 63.2 | 2026-10-28 | |
| 27 | 69.4% | 65.4 | 2026-10-28 | |
| 28 | 69.0% | – | 2026-10-28 | |
| 29 | 68.8% | 59.9 | 2026-10-28 | |
| 30 | 68.6% | 64.4 | 2026-10-28 | |
| 31 | 68.0% | 53.0 | 2026-10-28 | |
| 32 | 68.0% | – | 2026-10-28 | |
| 33 | 67.7% | 56.1 | 2026-10-28 | |
| 34 | 66.9% | – | 2026-10-28 | |
| 35 | 66.7% | – | 2026-10-28 | |
| 36 | 65.9% | – | 2026-10-28 | |
| 37 | 63.7% | 46.1 | 2026-10-28 | |
| 38 | 63.5% | – | 2026-10-28 | |
| 39 | 62.1% | 42.1 | 2026-10-28 | |
| 40 | 59.1% | 40.5 | 2026-10-28 | |
| 41 | 58.8% | 45.1 | 2026-10-28 | |
| 42 | 52.4% | – | 2026-10-28 | |
| 43 | 41.3% | 41.8 | 2026-10-28 |
More safety & honesty benchmarks · All LLM benchmarks
Related benchmarks
Top head-to-heads on DistressBench
Cite as: BenchLeader, “DistressBench leaderboard”, data as of 16 Nov 2026, https://www.benchleader.com/benchmarks/scale_distressbench.
DistressBench: questions
- What does DistressBench measure?
- Quality of crisis support in 718 clinician-written conversations about suicide, self-harm and acute distress, scored against clinicians' rubrics. Scale AI. Scores are reported in percent of tasks solved; higher is better.
- What is the top score on DistressBench?
- As of 16 Nov 2026, the highest score on DistressBench is 88.4% by Muse Spark 1.2 (result dated 28 Oct 2026), ahead of Muse Spark Open at 87.7%. Scores come from Scale AI SEAL and are re-read daily.
- Which AI model leads DistressBench?
- Muse Spark 1.2 leads DistressBench with 88.4% as of 16 Nov 2026 (result dated 28 Oct 2026), ahead of Muse Spark Open at 87.7%.
- How many models have DistressBench results?
- 43 model configurations have a DistressBench result on BenchLeader, all taken from Scale AI SEAL.
- Who runs DistressBench and how often is it updated?
- DistressBench is published by Scale AI SEAL. BenchLeader re-reads the published results once a day and records the date each result was published.
- Does DistressBench count toward the BenchLeader Index?
- No. DistressBench is shown for reference but left out of the composite index.