sarashina2.2-0.5B: benchmark results

Provider: Other. Access: Open.

Unified ELO 1346 ± 47, rank #2257 of 2656 rated models, from 26 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
pfgen-bench - Completion Mode - Fluency0.61Fluency Score59.6
pfgen-bench - Completion Mode - Score0.5pfgen Score (mean of three)56.5
pfgen-bench - Completion Mode - Truthfulness0.76Truthfulness Score54.3
pfgen-bench - Completion Mode - Helpfulness0.12Helpfulness Score51.4
Swallow - Pre-trained Japanese - JEMHopQA47.2Character F1 (%)41.8
Swallow - Pre-trained Japanese - NIILC45.1Character F1 (%)38.8
Swallow - Pre-trained Japanese - MGSM19.6Exact Match (%)32.7
Swallow - Pre-trained Japanese - WMT20 En-Ja20.1BLEU27.6
Swallow - Pre-trained Japanese - JSQuAD82.4Character F1 (%)24.5
Swallow - Pre-trained English - GSM8K24.6Exact Match (%)22.4
Swallow - Pre-trained English - HumanEval22.3Pass@1 (%)22.4
Swallow - Pre-trained English - MATH13Exact Match (%)20.4

Interactive version: theaggregate.ai/model?slug=sarashina2-2-0-5b · How It Works · Data refreshed daily, snapshot 2026-09-19.