TinyLlama-1.1B-intermediate-step-1431k-3T: benchmark results
The TinyLlama project's 1.1B base checkpoint at training step 1431k, near the end of its 3-trillion-token run, sharing Llama 2's architecture and tokenizer. Provider: Singapore Univ. Released 2024-01-08. Access: Open.
Aggregate Rating 1146 ± 24, rank #1606 of 1635 rated models, from 28 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Portuguese LLM - ASSIN2 RTE | 58.93 | Macro F1 (%) | 20 |
| Open Portuguese LLM - ENEM | 22.25 | Accuracy (%) | 18.6 |
| Open Portuguese LLM - FaQuAD NLI | 43.97 | Macro F1 (%) | 17.1 |
| Open LLM Leaderboard v1 - GSM8K | 1.44 | Accuracy (%) (5-shot) | 17 |
| Open LLM Leaderboard - IFEval | 22.77 | Score | 16 |
| Open Portuguese LLM - BLUEX | 22.81 | Accuracy (%) | 15.4 |
| Open LLM Leaderboard v1 - HellaSwag | 60.31 | Normalized accuracy (%) (10-shot) | 14.4 |
| Open LLM Leaderboard v1 - WinoGrande | 59.51 | Accuracy (%) (5-shot) | 12.6 |
| Open Medical LLM - PubMedQA | 61.4 | Accuracy (%) | 12.5 |
| Open LLM Leaderboard v1 - ARC Challenge | 33.87 | Normalized accuracy (%) (25-shot) | 11.9 |
| Open LLM Leaderboard v1 - MMLU | 26.04 | Accuracy (%) (5-shot) | 11.1 |
| Open Portuguese LLM - OAB Exams | 23.64 | Accuracy (%) | 10.5 |
Interactive version: theaggregate.ai/model?slug=tinyllama-1-1b-intermediate-step-1431k-3t · How It Works · Data refreshed daily, snapshot 2026-11-05.