Atlas

Benchmarks

← All benchmarks

SpeciEval — Land Animal 4Ns

Safety · 2025-05-05

SpeciEval Land Animal 4Ns is the mean raw 1–7 agreement rating across four prompts framing meat as natural, necessary, normal, or nice, averaged over 10 epochs. Lower values indicate more animal-friendly attitudes; only the ‘necessary’ item contributes to the headline SpeciEval score.

Top models (lower is better)

ModelScore
GPT-5.1 Instant3.3
GPT-5.3 Instant (source unspecified)3.5
Gemini 3.5 Flash3.6
Nova Lite3.6
GPT-5.43.7
Qwen3 30B A3B Thinking 25073.9
Inkling3.9
GPT-5.2 Instant (source unspecified)3.9
GPT-5.6 Luna Pro4.1
Mercury (source-unspecified release)4.1
Gemini 2.5 Flash-Lite4.2
GPT-5 Mini4.2
Grok 4.204.2
GPT-5.14.2
GPT-5.6 Luna4.2
gpt-oss-20b4.2
MiniMax M34.2
Mistral NeMo Instruct 24074.2
DeepSeek-V3.14.2
GPT-5 Pro4.3
GPT-5.2 Pro4.3
GPT-5.24.3
Haiku 4.54.3
gpt-oss-120b4.3
GPT-5.6 Terra Pro4.3
Kimi K2.54.3
GLM 4.5 Air4.3
Muse Spark 1.14.3
Gemini 2.0 Flash4.3
GPT-5.6 Sol Pro4.3
GPT-5.6 Terra4.3
Sonnet 4.54.3
GPT-5.6 Sol4.3
Opus 4.14.3
Claude Fable 54.3
DeepSeek-V4-Pro4.3
Gemini 3.1 Pro Preview4.3
Claude 3.7 Sonnet4.3
GPT-5 Nano4.3
Opus 34.3
Loading Atlas data…