Blur-7B-v1.2: benchmark results
liminerity's Blur-7b-v1.2 is a TIES merge of liminerity/Blured-Ties-7B, freecs/ThetaWave-7B. The released configuration uses 32 mistral decoder layers; merge recipe dtype is not a measured evaluator precision. Provider: liminerity. Access: Open.
Aggregate Rating 1423 ± 33, rank #5009 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 60.3 | MC2 (%) (0-shot) | 75.6 |
| Open LLM Leaderboard v1 - WinoGrande | 80.58 | Accuracy (%) (5-shot) | 74.7 |
| Open LLM Leaderboard v1 - ARC Challenge | 65.36 | Normalized accuracy (%) (25-shot) | 69.8 |
| Open LLM Leaderboard v1 - GSM8K | 52.84 | Accuracy (%) (5-shot) | 64.6 |
| Open LLM Leaderboard v1 - HellaSwag | 83.88 | Normalized accuracy (%) (10-shot) | 62.2 |
| Open LLM Leaderboard v1 - MMLU | 63.45 | Accuracy (%) (5-shot) | 60.8 |
Interactive version: theaggregate.ai/model?slug=blur-7b-v1-2 · How It Works · Data refreshed daily, snapshot 2026-11-08.