AA-Omniscience v1.1 Index
Indexes · 2026-10-27
AA-Omniscience v1.1 Index on 3,000 private questions across six economically relevant domains, revised from the 6,000-question v1 set by removing saturated and public-set questions, adding new questions and correcting ambiguous or erroneous ones. It rewards correct answers, penalizes hallucinations and does not penalize abstention, on a -100 to 100 scale. The changed question set is kept separate from v1.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 4 Argon | 46.1 |
| GPT-6 Astra | 44.6 |
| GPT-6.1 Sol | 42.4 |
| Grok 4.7 | 31.2 |
| Muse Wonder Full | 30.6 |
| Claude Sonnet 5.5 | 30.5 |
| Kimi K3.1 Pro | 29.9 |
| GPT-6.1 Terra | 29.6 |
| Muse Spark Open | 25.8 |
| Muse Wonder | 25.1 |
| Muse Spark 1.3 | 23.8 |
| Gemini 3.8 Flash | 23.7 |
| Step 5 Preview | 16.4 |
| Claude Haiku 5.5 | 12.3 |
| Qwen3.8 Max (0902) | 10.5 |
| Qwen3.8 2.4T A95B | 7.8 |
| Gemini 3.5 Flash-Lite | 6.9 |
| GLM-5.3 | 6.0 |
| Nemotron 3 Ultra 550B A55B | 3.4 |
| Beam 1.0 501B A23B | 2.3 |
| DeepSeek V4.1 Flash 1017 | 0.9 |
| GLM-5.3 Flash | 0.7 |
| Ling 3.1 Flash 1019 | -0.1 |
| Inkling | -1.5 |
| Kimi K3 | -1.7 |
| Command A+ (May 2026, BF16) | -2.7 |
| Nemotron 3.5 Super VL 120B A12B (precision unspecified) | -2.9 |
| MiMo-V2.6-Pro | -4.5 |
| K2 Horizon 375B A23B | -5.0 |
| MiniMax M3 | -5.5 |
| DeepSeek V4 Pro 0813 | -6.4 |
| Qwen3.8 27B | -6.4 |
| MiniMax M3.1 Flash | -6.4 |
| Mistral Large 4 Preview | -7.5 |
| GPT-6 Luna | -8.3 |
| GPT-5.6 Terra | -9.3 |
| Inkling-Small | -9.5 |
| MiMo-V2.6-Flash | -10.8 |
| Solar Mini 4 | -11.6 |
| Qwen3.8-Flash-Next | -12.2 |