MixtureofMerges-MoE-2x7bRP-v8: benchmark results
jsfs11 release with a Mixtral-architecture configuration: 2 local experts, 2 selected per token and 32 decoder layers. The hidden-gate merge recipe uses ChaoticNeutrals/RP_Vision_7B, ResplendentAI/DaturaCookie_7B. Expert labels do not independently establish total or active learned parameter counts. Provider: jsfs11. Access: Open.
Aggregate Rating 1454 ± 39, rank #3596 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - HellaSwag | 88.06 | Normalized accuracy (%) (10-shot) | 89.3 |
| Open LLM Leaderboard v1 - ARC Challenge | 71.33 | Normalized accuracy (%) (25-shot) | 89 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 68.69 | MC2 (%) (0-shot) | 87.9 |
| Open LLM Leaderboard v1 - WinoGrande | 82.95 | Accuracy (%) (5-shot) | 86.4 |
| Open LLM Leaderboard v1 - GSM8K | 64.52 | Accuracy (%) (5-shot) | 78.4 |
| Open LLM Leaderboard v1 - MMLU | 64.33 | Accuracy (%) (5-shot) | 68.4 |
Interactive version: theaggregate.ai/model?slug=mixtureofmerges-moe-2x7brp-v8 · How It Works · Data refreshed daily, snapshot 2026-11-08.