bilingual-gpt-neox-4B-8k: benchmark results
Rinna context-extension release of bilingual GPT-NeoX 4B, fine-tuned on 1.5B additional tokens using RoPE positional interpolation to extend its context from 2,048 to 8,192 tokens. Provider: rinna. Released 2023-07-31. Access: Open.
Aggregate Rating 1113 ± 38, rank #10592 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 47.48 | MC2 (%) (0-shot) | 37.9 |
| Open LLM Leaderboard v1 - HellaSwag | 43.94 | Normalized accuracy (%) (10-shot) | 8.7 |
| Open LLM Leaderboard v1 - MMLU | 25.38 | Accuracy (%) (5-shot) | 7.9 |
| Open LLM Leaderboard v1 - ARC Challenge | 28.58 | Normalized accuracy (%) (25-shot) | 7.8 |
| Open LLM Leaderboard v1 - GSM8K | 0 | Accuracy (%) (5-shot) | 4.5 |
| Open LLM Leaderboard v1 - WinoGrande | 47.99 | Accuracy (%) (5-shot) | 0.2 |
Interactive version: theaggregate.ai/model?slug=bilingual-gpt-neox-4b-8k · How It Works · Data refreshed daily, snapshot 2026-11-08.