SOLAR-10.7B-v1.0 — benchmark results

Upstage's Apache-2.0 10.7B pretrained base model (December 2023), built by depth up-scaling Mistral-7B weights on a Llama-2 architecture. Provider: Upstage. Released 2023-12-11. Access: Open.

Unified ELO 1418 ± 11, rank #1139 of 1776 rated models, from 321 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
EuroEval Romanian NLU - MultiWikiQA RO77.57Reading comprehension Score (%)99.5
EuroEval Portuguese NLU - MultiWikiQA PT79.82Reading comprehension Score (%)99.3
EuroEval Italian NLU - SQuAD IT77.48Reading comprehension Score (%)98.8
EuroEval Catalan NLU - MultiWikiQA CA78.2Reading comprehension Score (%)98.6
EuroEval Danish NLU - Multi Wiki QA DA82.34Reading comprehension Score (%)98.5
EuroEval Spanish NLU - MLQA ES70.14Reading comprehension Score (%)98.5
EuroEval Bulgarian NLU - MultiWikiQA BG79.17Reading comprehension Score (%)97.6
EuroEval Slovak NLU - Multi Wiki QA SK73.95Reading comprehension Score (%)97.5
EuroEval Polish NLU - PoQuAD63.61Reading comprehension Score (%)97
EuroEval Swedish NLU - Multi Wiki QA SV80.55Reading comprehension Score (%)96.4
EuroEval Slovene NLU - MultiWikiQA SL72.15Reading comprehension Score (%)96.1
EuroEval Hungarian NLU - MultiWikiQA HU69.79Reading comprehension Score (%)96

Interactive version: theaggregate.ai/model?slug=solar-10-7b-v1-0 · How the rankings work · Data refreshed daily, snapshot 2026-07-22.