NanoLlama-GQA-L10-A32_KV8-v13-KI: benchmark results
BEE-spoke-data release: A fine-tune of NanoLlama-GQA-L10-A32_KV8-v12-minipile with ten Llama decoder layers, 32 query heads and eight key/value heads. The generated card does not identify its training dataset reliably. Provider: BEE-spoke-data. Access: Open.
Aggregate Rating 1106 ± 40, rank #10645 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 44.77 | MC2 (%) (0-shot) | 28.5 |
| Open LLM Leaderboard v1 - GSM8K | 0.91 | Accuracy (%) (5-shot) | 14.6 |
| Open LLM Leaderboard v1 - MMLU | 25.37 | Accuracy (%) (5-shot) | 7.8 |
| Open LLM Leaderboard v1 - HellaSwag | 29.39 | Normalized accuracy (%) (10-shot) | 5.2 |
| Open LLM Leaderboard v1 - WinoGrande | 51.14 | Accuracy (%) (5-shot) | 5 |
| Open LLM Leaderboard v1 - ARC Challenge | 23.81 | Normalized accuracy (%) (25-shot) | 2.9 |
Interactive version: theaggregate.ai/model?slug=nanollama-gqa-l10-a32-kv8-v13-ki · How It Works · Data refreshed daily, snapshot 2026-11-08.