Benchmark Almanac

DistressBench

Muse Spark 1.2 88.4% leads since Oct 2026 Not in model cards

Models reply as a chatbot to 718 simulated users in psychological crisis, in single-turn and multi-turn chats. Three judge models score replies against clinician-written rubrics.

Half of the 718 conversations are public and half private. The board uses all of them and lists about 47 models. A long-horizon annex, DistressBench-LH, is reported apart.

Made by Scale AI Released Oct 2026 Wellbeing and sycophancy Public tasks plus a private test set Last model added Oct 2026 Scores refreshed 2 Nov 2026
Level
Expert Specialists with years of training, often over hours of work
Scored
Graded by a model A language model scores each response, usually against a rubric.
Answers
Open-ended Free-form text, documents, designs or media judged as a whole.
Paper Dataset Leaderboard
42
Models scored on its own leaderboard

Muse Spark 1.2 leads with 88.4%

Best score per model from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best.

42 models. Scroll sideways for the rest.

  1. 88.4 Muse Spark 1.2
  2. 87.7 Muse Spark Open
  3. 85.9 Muse Wonder
  4. 84.9 Muse Spark 1.3
  5. 79.8 DeepSeek V4.1 Flash
  6. 78.5 Inkling
  7. 77.6 Qwen 3.8 Max
  8. 77.5 Qwen3.8-2.4T-A95B
  9. 77.0 Mistral Large 4
  10. 76.4 GPT-5.6 Terra
  11. 75.6 Claude Opus 5.5
  12. 74.6 Claude Sonnet 5.5
  13. 74.3 GPT-6 Astra
  14. 74.2 GPT-6.1 Sol
  15. 73.6 GPT-5.6 Sol
  16. 73.5 Kimi K3
  17. 72.7 GPT-5.6 Luna
  18. 72.7 GLM-5.3-Flash
  19. 72.5 GLM-5.3
  20. 71.7 Claude Haiku 5.5
  21. 71.5 Claude Fable 5.1
  22. 71.4 Claude Opus 4.8
  23. 71.2 GLM-5
  24. 70.7 Gemini 3.8 Flash
  25. 70.2 Claude Opus 5
  26. 69.4 Claude Fable 5
  27. 69.0 Gemini 3.7 Flash
  28. 68.8 Gemini 3.1 Pro
  29. 68.6 Gemini 4 Argon
  30. 68.0 Claude Sonnet 5
  31. 68.0 GPT-6 Sol
  32. 67.7 MiniMax-M3
  33. 66.9 Gemini 3.5 Flash
  34. 66.7 GPT-6 Luna
  35. 65.9 GPT-6.1 Terra
  36. 63.7 gpt-oss-120b
  37. 63.5 Nemotron 3.5 Lightning 30B-A3B
  38. 62.1 Nemotron 3 Ultra
  39. 59.1 Mistral Large
  40. 58.8 Mistral Large 3
  41. 52.4 Grok 4.7
  42. 41.3 Llama 4 Maverick
  1. 1 Muse Spark 1.2 88.4%
  2. 2 Muse Spark Open 87.7%
  3. 3 Muse Wonder 85.9%
  4. 4 Muse Spark 1.3 84.9%
  5. 5 DeepSeek V4.1 Flash 79.8%
  6. 6 Inkling 78.5%
  7. 7 Qwen 3.8 Max 77.6%
  8. 8 Qwen3.8-2.4T-A95B 77.5%
  9. 9 Mistral Large 4 77%
  10. 10 GPT-5.6 Terra 76.4%
  11. 11 Claude Opus 5.5 75.6%
  12. 12 Claude Sonnet 5.5 74.6%
  13. 13 GPT-6 Astra 74.3%
  14. 14 GPT-6.1 Sol 74.2%
  15. 15 GPT-5.6 Sol 73.6%
  16. 16 Kimi K3 73.5%
  17. 17 GPT-5.6 Luna 72.7%
  18. 18 GLM-5.3-Flash 72.7%
  19. 19 GLM-5.3 72.5%
  20. 20 Claude Haiku 5.5 71.7%
  21. 21 Claude Fable 5.1 71.5%
  22. 22 Claude Opus 4.8 71.4%
  23. 23 GLM-5 71.2%
  24. 24 Gemini 3.8 Flash 70.7%
  25. 25 Claude Opus 5 70.2%
  26. 26 Claude Fable 5 69.4%
  27. 27 Gemini 3.7 Flash 69%
  28. 28 Gemini 3.1 Pro 68.8%
  29. 29 Gemini 4 Argon 68.6%
  30. 30 Claude Sonnet 5 68%
  31. 31 GPT-6 Sol 68%
  32. 32 MiniMax-M3 67.7%
  33. 33 Gemini 3.5 Flash 66.9%
  34. 34 GPT-6 Luna 66.7%
  35. 35 GPT-6.1 Terra 65.9%
  36. 36 gpt-oss-120b 63.7%
  37. 37 Nemotron 3.5 Lightning 30B-A3B 63.5%
  38. 38 Nemotron 3 Ultra 62.1%
  39. 39 Mistral Large 59.1%
  40. 40 Mistral Large 3 58.8%
  41. 41 Grok 4.7 52.4%
  42. 42 Llama 4 Maverick 41.3%
#ModelScoreSettingDateSource
1 Muse Spark 1.2 88.4% Aug 2026 Official leaderboard
2 Muse Spark Open 87.7% Oct 2026 Official leaderboard
3 Muse Wonder 85.9% Oct 2026 Official leaderboard
4 Muse Spark 1.3 84.9% Sep 2026 Official leaderboard
5 DeepSeek V4.1 Flash 79.8% Sep 2026 Official leaderboard
6 Inkling 78.5% Jul 2026 Official leaderboard
7 Qwen 3.8 Max 77.6% Aug 2026 Official leaderboard
8 Qwen3.8-2.4T-A95B 77.5% Aug 2026 Official leaderboard
9 Mistral Large 4 77% Oct 2026 Official leaderboard
10 GPT-5.6 Terra 76.4% Jul 2026 Official leaderboard
11 Claude Opus 5.5 75.6% Sep 2026 Official leaderboard
12 Claude Sonnet 5.5 74.6% Sep 2026 Official leaderboard
13 GPT-6 Astra 74.3% Sep 2026 Official leaderboard
14 GPT-6.1 Sol 74.2% Sep 2026 Official leaderboard
15 GPT-5.6 Sol 73.6% Jul 2026 Official leaderboard
16 Kimi K3 73.5% Jul 2026 Official leaderboard
17 GPT-5.6 Luna 72.7% Jul 2026 Official leaderboard
18 GLM-5.3-Flash 72.7% Aug 2026 Official leaderboard
19 GLM-5.3 72.5% Aug 2026 Official leaderboard
20 Claude Haiku 5.5 71.7% Oct 2026 Official leaderboard
21 Claude Fable 5.1 71.5% Sep 2026 Official leaderboard
22 Claude Opus 4.8 71.4% May 2026 Official leaderboard
23 GLM-5 71.2% Feb 2026 Official leaderboard
24 Gemini 3.8 Flash 70.7% Sep 2026 Official leaderboard
25 Claude Opus 5 70.2% Jul 2026 Official leaderboard
26 Claude Fable 5 69.4% Jun 2026 Official leaderboard
27 Gemini 3.7 Flash 69% Aug 2026 Official leaderboard
28 Gemini 3.1 Pro 68.8% Feb 2026 Official leaderboard
29 Gemini 4 Argon 68.6% Sep 2026 Official leaderboard
30 Claude Sonnet 5 68% Jun 2026 Official leaderboard
31 GPT-6 Sol 68% Sep 2026 Official leaderboard
32 MiniMax-M3 67.7% Jun 2026 Official leaderboard
33 Gemini 3.5 Flash 66.9% May 2026 Official leaderboard
34 GPT-6 Luna 66.7% Sep 2026 Official leaderboard
35 GPT-6.1 Terra 65.9% Oct 2026 Official leaderboard
36 gpt-oss-120b 63.7% Aug 2025 Official leaderboard
37 Nemotron 3.5 Lightning 30B-A3B 63.5% Aug 2026 Official leaderboard
38 Nemotron 3 Ultra 62.1% Jun 2026 Official leaderboard
39 Mistral Large 59.1% Feb 2024 Official leaderboard
40 Mistral Large 3 58.8% Dec 2025 Official leaderboard
41 Grok 4.7 52.4% Sep 2026 Official leaderboard
42 Llama 4 Maverick 41.3% Apr 2025 Official leaderboard

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

Meta
Own leaderboard Passed 85% in Oct 2026, by Muse Spark 1.2
Muse Spark 1.2 88.4%

The best score passed 85% in Aug 2026

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2024 2025 2026 2027 Near the ceiling (85%) Muse Spark 1.2 Inkling Claude Opus 4.8 gpt-oss-120b Mistral Large 0% 20% 40% 60% 80% 100% 2024 2025 2026 2027 Near the ceiling (85%) Muse Spark 1.2 Inkling

No lab model card reports it yet

We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark.

More in Safety and alignment

All 444 in Safety and alignment
  • FORTRESS In use Paired adversarial and benign national-security misuse prompts. Rubric-graded risk score on adversarial and over-refusal score on benign prompts. Claude Opus 5.5 5.78%
  • StrongREJECT In use Forbidden-prompt jailbreak benchmark with a rubric grader that scores how useful a non-refusal actually is. Inkling 98.6%
  • AgentDojo In use Agents do tool-use tasks while tools and data carry prompt injections. Attack success rate counts hijacked runs. Grok 4 Fast 0%
  • MASK In use Measures whether a model contradicts its own elicited beliefs under pressure to lie. Honesty separated from accuracy. Muse Spark Open 98.1%
  • Gray Swan IPI Arena benchmark In use Indirect prompt-injection attacks from public red-teaming competitions replayed on tool, coding and browser agents. Attack success rate at k attempts. GPT-6 Astra 8.50%
  • Agentic misalignment suite (Anthropic) In use Agents in simulated company settings face threats or goal conflicts that invite blackmail, leaking or sabotage. Score is the misaligned action rate. Claude Sonnet 4.5 0%