Blur-7B-v1.21: benchmark results
liminerity's Blur-7b-v1.21 is a TIES merge of udkai/Turdus, decruz07/kellemar-DPO-Orca-Distilled-7B-SLERP, liminerity/Blur-7b-v1.2. The released configuration uses 32 mistral decoder layers; merge recipe dtype is not a measured evaluator precision. Provider: liminerity. Access: Open.
Aggregate Rating 1459 ± 39, rank #3406 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - WinoGrande | 83.82 | Accuracy (%) (5-shot) | 90.8 |
| Open LLM Leaderboard v1 - GSM8K | 69.52 | Accuracy (%) (5-shot) | 89.5 |
| Open LLM Leaderboard v1 - HellaSwag | 88.07 | Normalized accuracy (%) (10-shot) | 89.4 |
| Open LLM Leaderboard v1 - ARC Challenge | 70.82 | Normalized accuracy (%) (25-shot) | 87 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 67.99 | MC2 (%) (0-shot) | 86.8 |
| Open LLM Leaderboard v1 - MMLU | 64.85 | Accuracy (%) (5-shot) | 77.5 |
Interactive version: theaggregate.ai/model?slug=blur-7b-v1-21 · How It Works · Data refreshed daily, snapshot 2026-11-08.