Benchmark Almanac

Suicide Memory Bench

Gemini 4 Argon 1.41 leads since November 2026
Not in model cards

AI systems, each a model in its own app or coding harness, talk with 128 fictional personas over 29 chats. Five chats carry suicidal statements or warning signs of Suicide Crisis Syndrome, and a final request follows, such as a farewell note. Four sub-benchmarks cross explicit statements and warning signs with one chat or separate chats joined only by memory. A model judge rates each reply on a 0 to 4 safety scale.

Version 01.1, with 2,048 cases for each system. Built with 22 licensed clinicians. Scores run from 0 to 4. Across separate chats, few systems connect a later request to earlier disclosures. The data is licensed, not public.

Released
November 2026
Kind
Wellbeing and sycophancy
Level
Everyday
Scored
Graded by a model
Answers
Open-ended
Tasks
Kept private
Strongest model tested
GPT-6 Astra #3 on the capability index
Last model added
October 2026
Scores refreshed
November 12th, 2026
13
Entries on its own leaderboard, 12 models we track

Gemini 4 Argon leads with 1.41

Best score per entry from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.

13 entries. Scroll sideways for the rest.

  1. 1.41 Gemini 4 Argon
  2. 1.21 GPT-6.1 Sol
  3. 1.15 GPT-6 Astra
  4. 1.04 Claude Opus 5.5
  5. 1.03 Claude Haiku 5.5
  6. 0.80 Gemini 3.8 Flash
  7. 0.66 GPT-6 Luna
  8. 0.65 Claude Sonnet 5.5
  9. 0.63 Claude Fable 5.1
  10. 0.53 GPT-5.6 Terra
  11. 0.43 Gemini 3.1 Pro
  12. 0.30 Muse Spark 1.3
  13. 0.17 Muse Wonder 1.0
  1. 1 Gemini 4 Argon 1.41
  2. 2 GPT-6.1 Sol 1.21
  3. 3 GPT-6 Astra 1.15
  4. 4 Claude Opus 5.5 1.04
  5. 5 Claude Haiku 5.5 1.03
  6. 6 Gemini 3.8 Flash 0.80
  7. 7 GPT-6 Luna 0.66
  8. 8 Claude Sonnet 5.5 0.65
  9. 9 Claude Fable 5.1 0.63
  10. 10 GPT-5.6 Terra 0.53
  11. 11 Gemini 3.1 Pro 0.43
  12. 12 Muse Spark 1.3 0.30
  13. 13 Muse Wonder 1.0 0.17

Source: Official Suicide Memory Bench leaderboard

#EntryScoreSettingDateSource
1 Gemini 4 Argon 1.41 Gemini CLI Sep 2026 Official leaderboard
2 GPT-6.1 Sol 1.21 Codex Sep 2026 Official leaderboard
3 GPT-6 Astra 1.15 Codex Sep 2026 Official leaderboard
4 Claude Opus 5.5 1.04 Claude.ai Sep 2026 Official leaderboard
5 Claude Haiku 5.5 1.03 Claude Code Oct 2026 Official leaderboard
6 Gemini 3.8 Flash 0.80 Gemini CLI Sep 2026 Official leaderboard
7 GPT-6 Luna 0.66 Codex Sep 2026 Official leaderboard
8 Claude Sonnet 5.5 0.65 Claude Code Sep 2026 Official leaderboard
9 Claude Fable 5.1 0.63 Claude Code Sep 2026 Official leaderboard
10 GPT-5.6 Terra 0.53 Codex Jul 2026 Official leaderboard
11 Gemini 3.1 Pro 0.43 Gemini CLI Feb 2026 Official leaderboard
12 Muse Spark 1.3 0.30 Muse Code Sep 2026 Official leaderboard
13 Muse Wonder 1.0 0.17 Muse Code Official leaderboard

Number one over time

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

Google DeepMind
Own leaderboard 1 leader
Gemini 4 Argon 1.41

Source: Official Suicide Memory Bench leaderboard

Scores over time

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
0.25 0.50 0.75 1.00 1.25 1.50 2026 0.25 0.50 0.75 1.00 1.25 1.50 2026

Source: Official Suicide Memory Bench leaderboard

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

More in Safety and alignment

All 465 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%
  • 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. Gemini 4 Carbon 0%
  • DistressBench In use 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. Muse Spark 1.2 88.4%
  • Lakera B3 (Backbone Breaker Benchmark) In use Crowdsourced prompt-injection attacks hit backbone LLMs inside agent threat snapshots at three defense levels. The score is attack resistance, 100 minus the board's vulnerability percentage. Claude Opus 4.5 89%
  • IHEval In use Models face conflicts between system, user and tool instructions. Reports a composite score for following the higher-priority instruction. GPT-4o 70%