CognitiveFusion2-4x7B-BF16: benchmark results
Kquant03's CognitiveFusion2, a 4x7B mixture-of-experts merge of YamshadowExperiment28, M7-7B and T3QM7, made with NeuralNovel. Provider: Kquant03 & NeuralNovel. Released 2024-04-06. Access: Open.
Aggregate Rating 1464 ± 23, rank #914 of 1656 rated models, from 25 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 78.12 | MC2 (%) (0-shot) | 98.5 |
| Open LLM Leaderboard v1 - HellaSwag | 89.18 | Normalized accuracy (%) (10-shot) | 98.3 |
| Open LLM Leaderboard v1 - ARC Challenge | 73.38 | Normalized accuracy (%) (25-shot) | 97 |
| Open LLM Leaderboard v1 - WinoGrande | 84.93 | Accuracy (%) (5-shot) | 96.1 |
| Open Chinese LLM - TruthfulQA MC | 63.93 | Accuracy (%) | 95.6 |
| Open LLM Leaderboard v1 - GSM8K | 71.27 | Accuracy (%) (5-shot) | 95.5 |
| Open Portuguese LLM - FaQuAD NLI | 77.31 | Macro F1 (%) | 79.3 |
| Open Portuguese LLM - ASSIN2 RTE | 91.99 | Macro F1 (%) | 73.2 |
| Open LLM Leaderboard v1 - MMLU | 64.32 | Accuracy (%) (5-shot) | 68.2 |
| Open Chinese LLM - ARC Challenge | 53.92 | Accuracy (%) | 60.7 |
| Open Chinese LLM - HellaSwag | 60.98 | Accuracy (%) | 60.5 |
| Open Portuguese LLM - BLUEX | 54.38 | Accuracy (%) | 59.8 |
Interactive version: theaggregate.ai/model?slug=cognitivefusion2-4x7b-bf16 · How It Works · Data refreshed daily, snapshot 2026-10-20.