DeciLM-7B: benchmark results
Deci AI's 7B base model built with its AutoNAC neural architecture search, using variable grouped-query attention for higher throughput and an 8K-token context. Provider: Deci AI. Released 2023-12-10. Access: Open.
Aggregate Rating 1365 ± 26, rank #1350 of 1656 rated models, from 28 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Portuguese LLM - FaQuAD NLI | 78.57 | Macro F1 (%) | 84.6 |
| Open CoT - LSAT Analytical Reasoning | 6.52 | CoT Gain (%) | 81.3 |
| Open LLM Leaderboard - MuSR | 43.59 | Score | 75.1 |
| Open LLM Leaderboard v1 - WinoGrande | 79.95 | Accuracy (%) (5-shot) | 71.6 |
| Open Portuguese LLM - ASSIN2 RTE | 91.23 | Macro F1 (%) | 65.4 |
| Open LLM Leaderboard v1 - GSM8K | 47.38 | Accuracy (%) (5-shot) | 60.8 |
| Open LLM Leaderboard v1 - HellaSwag | 82.51 | Normalized accuracy (%) (10-shot) | 51.8 |
| Open LLM Leaderboard - GPQA | 29.53 | Score | 50.7 |
| Open LLM Leaderboard v1 - MMLU | 59.76 | Accuracy (%) (5-shot) | 45.8 |
| Open LLM Leaderboard v1 - ARC Challenge | 59.39 | Normalized accuracy (%) (25-shot) | 43.2 |
| Open Portuguese LLM - ENEM | 54.23 | Accuracy (%) | 37.3 |
| Open CoT - LSAT Reading Comprehension | 8.18 | CoT Gain (%) | 34.4 |
Interactive version: theaggregate.ai/model?slug=decilm-7b · How It Works · Data refreshed daily, snapshot 2026-10-20.