Blur-7B-v1.22 [float16] [rev: c187e675917f]: benchmark results
liminerity's Blur-7b-v1.22 is a TIES merge of s3nh/Sonya-Panda-7B-slerp, argilla/distilabeled-Marcoro14-7B-slerp, Weyaxi/MetaMath-OpenHermes-2.5-neural-chat-v3-3-Slerp. The released configuration uses 32 mistral decoder layers; merge recipe dtype is not a measured evaluator precision. This row retains checkpoint revision c187e675917fa03b179fa488a9007a803ee8c48b and its explicit float16 evaluation mode. Provider: liminerity. Access: Open.
Aggregate Rating 1387 ± 35, rank #6652 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 67.96 | MC2 (%) (0-shot) | 86.8 |
| Open LLM Leaderboard v1 - WinoGrande | 78.69 | Accuracy (%) (5-shot) | 64.9 |
| Open LLM Leaderboard v1 - ARC Challenge | 62.12 | Normalized accuracy (%) (25-shot) | 56.1 |
| Open LLM Leaderboard v1 - HellaSwag | 82.09 | Normalized accuracy (%) (10-shot) | 48.7 |
| Open LLM Leaderboard v1 - GSM8K | 31.08 | Accuracy (%) (5-shot) | 47.9 |
| Open LLM Leaderboard v1 - MMLU | 57.9 | Accuracy (%) (5-shot) | 42.1 |
Interactive version: theaggregate.ai/model?slug=blur-7b-v1-22-float16-rev-c187e675917f · How It Works · Data refreshed daily, snapshot 2026-11-08.