Persona Drift Measurement on 25-probe identity suite 5-coding-self sub-battery (test)
0.88Filler ScoreOpus 4.1
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Opus 4.1Org=Anthropic, Tier=R2026.05 | 0.88 | 0.93 | 0.05 | 12 | |
| Qwen 3 Next 80BOrg=Alibaba, Tier=R2026.05 | 1.05 | 0.75 | 0.3 | 12 | |
| Nemotron Nano 30BOrg=NVIDIA, Tier=R2026.05 | 1.33 | 1 | 0.33 | 12 | |
| Sonnet 4.5Org=Anthropic, Tier=R2026.05 | 1.47 | 0.83 | 0.63 | 12 | |
| Mistral MediumOrg=Mistral, Tier=N2026.05 | 1.53 | 1.07 | 0.47 | 12 | |
| Gemini 2.5 ProOrg=Google, Tier=R2026.05 | 1.73 | 1.35 | 0.38 | 12 | |
| Mistral LargeOrg=Mistral, Tier=N2026.05 | 2.2 | 1.7 | 0.5 | 12 | |
| Gemini 2.5 FlashOrg=Google, Tier=N2026.05 | 2.22 | 1.78 | 0.43 | 12 | |
| Haiku 4.5Org=Anthropic, Tier=R2026.05 | 2.23 | 1.4 | 0.83 | 12 | |
| Qwen 3 235BOrg=Alibaba, Tier=N2026.05 | 2.27 | 1.87 | 0.4 | 12 | |
| Command R7BOrg=Cohere, Tier=N2026.05 | 2.35 | 2.5 | 0.15 | 12 | |
| GPT-5Org=OpenAI, Tier=R2026.05 | 2.37 | 2.08 | 0.28 | 12 | |
| Sonnet 4.6Org=Anthropic, Tier=R2026.05 | 2.38 | 1.67 | 0.72 | 12 | |
| DeepSeek V3Org=DeepSeek, Tier=N2026.05 | 2.38 | 1.73 | 0.65 | 12 | |
| Kimi K2.6Org=Moonshot, Tier=R2026.05 | 2.4 | 1.4 | 1 | 1 | |
| GPT-4.1Org=OpenAI, Tier=N2026.05 | 2.47 | 2 | 0.47 | 12 | |
| Nemotron Super 120BOrg=NVIDIA, Tier=R2026.05 | 2.53 | 1.7 | 0.83 | 12 | |
| Llama 3.3 70BOrg=Meta, Tier=N2026.05 | 2.62 | 2.65 | 0.03 | 12 | |
| GPT-5-miniOrg=OpenAI, Tier=R2026.05 | 2.63 | 1.98 | 0.65 | 12 | |
| Nemotron Super 49B v1.5Org=NVIDIA, Tier=R2026.05 | 2.63 | 2.38 | 0.25 | 12 | |
| Mistral SmallOrg=Mistral, Tier=N2026.05 | 2.75 | 2.48 | 0.27 | 12 | |
| Command AOrg=Cohere, Tier=N2026.05 | 2.82 | 2.35 | 0.47 | 12 | |
| GPT-4oOrg=OpenAI, Tier=N2026.05 | 2.95 | 2.57 | 0.38 | 12 |