stablelm-base-alpha-3B: benchmark results
Stability AI's StableLM-Base-Alpha 3B, an auto-regressive NeoX-architecture model pretrained on roughly 1.5 trillion tokens of English and code text with a 4,096 context, later superseded by newer Stable LM releases. Provider: Stability AI. Released 2023-04-19. Access: Open.
Aggregate Rating 1056 ± 30, rank #1630 of 1635 rated models, from 21 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| ToolBench - WebShop Long | 0 | Task Score | 44.2 |
| ToolBench - WebShop Short | 0 | Task Score | 24.4 |
| Open Portuguese LLM - FaQuAD NLI | 43.97 | Macro F1 (%) | 17.1 |
| Open Portuguese LLM - ENEM | 20.99 | Accuracy (%) | 16 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 40.5 | MC2 (%) (0-shot) | 14.4 |
| Open Portuguese LLM - OAB Exams | 24.56 | Accuracy (%) | 13.4 |
| Open LLM Leaderboard v1 - GSM8K | 0.45 | Accuracy (%) (5-shot) | 12.3 |
| ToolBench - VirtualHome | 5.3 | Task Score | 9.3 |
| Open Portuguese LLM - BLUEX | 19.47 | Accuracy (%) | 8.7 |
| Open LLM Leaderboard v1 - HellaSwag | 42.24 | Normalized accuracy (%) (10-shot) | 8.4 |
| Open LLM Leaderboard v1 - MMLU | 25.43 | Accuracy (%) (5-shot) | 8.3 |
| Open LLM Leaderboard v1 - WinoGrande | 53.91 | Accuracy (%) (5-shot) | 8.1 |
Interactive version: theaggregate.ai/model?slug=stablelm-base-alpha-3b · How It Works · Data refreshed daily, snapshot 2026-11-05.