bilingual-gpt-neox-4B-instruction-sft: benchmark results
Rinna supervised instruction-tuned bilingual GPT-NeoX release. The publisher replaced the July 31, 2023 checkpoint with a newly trained MIT-licensed checkpoint on August 2, 2023; the captured OpenLLM evaluator revision uses the replacement release. Provider: rinna. Released 2023-08-02. Access: Open.
Aggregate Rating 1121 ± 40, rank #10502 of 10916 rated models, from 10 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| pfgen-bench - Completion Mode - Truthfulness | 0.65 | Truthfulness Score | 37.7 |
| pfgen-bench - Completion Mode - Score | 0.38 | pfgen Score (mean of three) | 29.7 |
| pfgen-bench - Completion Mode - Fluency | 0.46 | Fluency Score | 27 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 43.76 | MC2 (%) (0-shot) | 24.7 |
| pfgen-bench - Completion Mode - Helpfulness | 0.04 | Helpfulness Score | 23 |
| Open LLM Leaderboard v1 - HellaSwag | 47.5 | Normalized accuracy (%) (10-shot) | 9.5 |
| Open LLM Leaderboard v1 - ARC Challenge | 28.07 | Normalized accuracy (%) (25-shot) | 7.3 |
| Open LLM Leaderboard v1 - WinoGrande | 52.33 | Accuracy (%) (5-shot) | 6.6 |
| Open LLM Leaderboard v1 - GSM8K | 0 | Accuracy (%) (5-shot) | 4.5 |
| Open LLM Leaderboard v1 - MMLU | 23.12 | Accuracy (%) (5-shot) | 1 |
Interactive version: theaggregate.ai/model?slug=bilingual-gpt-neox-4b-instruction-sft · How It Works · Data refreshed daily, snapshot 2026-11-08.