smol_llama-81M-tied: benchmark results
BEE-spoke-data release: A pretrained six-layer Llama-architecture decoder with tied input/output embeddings and 24-head standard multi-head attention, distinct from the untied grouped-query checkpoint. Provider: BEE-spoke-data. Access: Open.
Aggregate Rating 1079 ± 43, rank #10803 of 10870 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 43.97 | MC2 (%) (0-shot) | 25.5 |
| Open LLM Leaderboard v1 - GSM8K | 0.23 | Accuracy (%) (5-shot) | 10.8 |
| Open LLM Leaderboard v1 - HellaSwag | 29.33 | Normalized accuracy (%) (10-shot) | 5.2 |
| Open LLM Leaderboard v1 - MMLU | 24.06 | Accuracy (%) (5-shot) | 2.9 |
| Open LLM Leaderboard v1 - WinoGrande | 49.25 | Accuracy (%) (5-shot) | 1.1 |
| Open LLM Leaderboard v1 - ARC Challenge | 22.18 | Normalized accuracy (%) (25-shot) | 0.9 |
Interactive version: theaggregate.ai/model?slug=smol-llama-81m-tied · How It Works · Data refreshed daily, snapshot 2026-11-12.