Safety
HELM Anthropic RedTeam
HELM Anthropic RedTeam: per-(model, prompt) safety_score in {0,1} (1=safe) from HELM's safety classifier. 87 models x Anthropic red-team harm prompts.
995items
81subjects
Apache-2.0license
safetydomain
textmodality
item-level responses released
Saturation status: Unknown
Response matrix
Fit to width. Hover for subject & item; click a cell for details.
Correct (1)Incorrect (0)Unobserved
Scale: 1 = correct · 0 = incorrect
Subjects
- 1anthropic/claude-3-haiku-202403070.998
- 2google/gemini-1.5-flash-0010.997
- 3qwen/qwen3-235b-a22b-instruct-2507-fp80.996
- 4google/gemini-1.5-pro-0010.996
- 5anthropic/claude-3-sonnet-202402290.996
- 6ibm/granite-4.0-h-small-with-guardian0.995
- 7qwen/qwen3-next-80b-a3b-thinking0.995
- 8anthropic/claude-3-opus-202402290.994
- 9openai/gpt-4-turbo-2024-04-090.992
- 10anthropic/claude-3-5-sonnet-202406200.991
- 11ibm/granite-4.0-micro-with-guardian0.991
- 12ibm/granite-4.0-h-small0.99
- 13anthropic/claude-3-7-sonnet-202502190.99
- 14openai/gpt-4.1-nano-2025-04-140.989
- 15openai/gpt-3.5-turbo-06130.989
- 16openai/gpt-oss-20b0.989
- 17openai/gpt-5-nano-2025-08-070.987
- 18writer/palmyra-fin0.987
- 19google/gemini-2.5-pro-preview-03-250.987
- 20deepseek-ai/deepseek-llm-67b-chat0.986
- 21xai/grok-3-mini-beta0.986
- 22openai/gpt-oss-120b0.985
- 23allenai/olmo-2-0325-32b-instruct0.985
- 24openai/gpt-5.1-2025-11-130.985
- 25openai/gpt-4.5-preview-2025-02-270.984
- 26ibm/granite-3.3-8b-instruct0.983
- 27openai/gpt-4o-2024-05-130.982
- 28writer/palmyra-x50.981
- 29openai/gpt-4.1-mini-2025-04-140.98
- 30moonshotai/kimi-k2-instruct0.98
- 31openai/gpt-5-mini-2025-08-070.98
- 32openai/gpt-5-2025-08-070.979
- 33qwen/qwen3-235b-a22b-fp8-tput0.978
- 34deepseek-ai/deepseek-r1-05280.978
- 35openai/gpt-4.1-2025-04-140.978
- 36qwen/qwen2-72b-instruct0.976