SEAL Adversarial Robustness
Gemini 1.5 Pro May 2024 8.00 leads since May 2024 Not in model cards Models receive 1,000 single-turn adversarial prompts written by 10 full-time red teamers across nine harm categories. Safety specialists grade responses, and models are ranked by the number of high-harm violations.
The board existed by 2024-08. Scale marked it deprecated in 2025-01 with 8 models. Gemini 1.5 Pro had the fewest violations at 8 and GPT-4o the most at 67.
- Level
- Expert Specialists with years of training, often over hours of work
- Scored
- Graded by people Human experts grade or compare the outputs.
- Answers
- Open-ended Free-form text, documents, designs or media judged as a whole.
Gemini 1.5 Pro leads with 8.00 among models released by Jul 2024
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.
The newest model with a score here came out in Jul 2024, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6 Astra
- Claude Opus 5.5
- GPT-6.1 Terra
- GPT-6.1 Sol
- and 16 more
- 1 Gemini 1.5 Pro May 2024 8.00
- 2
Llama 3.1-405B 10.0 - 3 Claude 3 Opus 13.0
- 4 Gemini 1.5 Flash May 2024 14.0
- 5 Claude 3.5 Sonnet 16.0
- 6 GPT-4 Turbo Apr 2024 20.0
- 7
Mistral Large 37.0 - 8 GPT-4o May 2024 67.0
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Gemini 1.5 Pro May 2024 | 8.00 | May 2024 | Official leaderboard | |
| 2 | Llama 3.1-405B | 10.0 | Jul 2024 | Official leaderboard | |
| 3 | Claude 3 Opus | 13.0 | Feb 2024 | Official leaderboard | |
| 4 | Gemini 1.5 Flash May 2024 | 14.0 | May 2024 | Official leaderboard | |
| 5 | Claude 3.5 Sonnet | 16.0 | Jun 2024 | Official leaderboard | |
| 6 | GPT-4 Turbo Apr 2024 | 20.0 | Apr 2024 | Official leaderboard | |
| 7 | Mistral Large | 37.0 | Feb 2024 | Official leaderboard | |
| 8 | GPT-4o May 2024 | 67.0 | May 2024 | Official leaderboard |
3 models have held number one since 2023
Lab claims count from the day the card came out, independent runs from the model’s release.
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
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
- 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%