DeepSeek-R1-ReDistill-Llama3-8B-v1.1 Benchmark Results

DeepSeek-R1-ReDistill-Llama3-8B-v1.1 — benchmark results

Provider: DeepSeek. Released 2025-01-29. Access: Open.

Unified ELO 1331 ± 13, rank #1471 of 1776 rated models, from 179 benchmark results.

Benchmark	Score	Metric	Percentile
Open LLM Leaderboard - MATH Level 5	32.85	Score	83.7
Open Japanese LLM - Wiki NER SET F1	5.31	Score (%)	51.2
Open LLM Leaderboard - IFEval	37.04	Score	36
Open Japanese LLM - Wiki PAS SET F1	4.72	Score (%)	34.1
Open Japanese LLM - Jsem Exact Match	72.66	Score (%)	33.7
Open Japanese LLM - Jnli Exact Match	72.93	Score (%)	33.4
Open Arabic LLM - Arabic MMLU HT Econometrics	26.32	Accuracy (%)	32.7
Open Japanese LLM - Xlsum JA Bert Score JA F1	67.61	Score (%)	31.5
Open Japanese LLM - Wikicorpus J TO E Bert Score EN F1	88.88	Score (%)	30.5
Open Japanese LLM - CG	0.4	Score (%)	30
Open Japanese LLM - Jamp Exact Match	50.86	Score (%)	29.9
Open Arabic LLM - Arabic MMLU HT Abstract Algebra	25	Accuracy (%)	29.6