UNAversal-2x7B-v1: benchmark results
fblgit experimental UNA mixture of experts derived from Intel neural-chat, with two MLP experts per decoder layer. Provider: fblgit. Access: Open.
Aggregate Rating 1455 ± 42, rank #3552 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - ARC Challenge | 73.38 | Normalized accuracy (%) (25-shot) | 97 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 69.93 | MC2 (%) (0-shot) | 89.7 |
| Open LLM Leaderboard v1 - HellaSwag | 87.87 | Normalized accuracy (%) (10-shot) | 88.4 |
| Open LLM Leaderboard v1 - GSM8K | 67.55 | Accuracy (%) (5-shot) | 83.8 |
| Open LLM Leaderboard v1 - WinoGrande | 82.08 | Accuracy (%) (5-shot) | 82.2 |
| Open LLM Leaderboard v1 - MMLU | 63.49 | Accuracy (%) (5-shot) | 61.1 |
Interactive version: theaggregate.ai/model?slug=unaversal-2x7b-v1 · How It Works · Data refreshed daily, snapshot 2026-11-08.