SOLAR-10.7B-Instruct-v1.0 — benchmark results
Upstage's 10.7B instruct tune of its depth-upscaled SOLAR base (Mistral 7B weights in a Llama 2-style stack), aligned with SFT and DPO (December 2023). Provider: Upstage. Released 2023-12-13. Access: Open.
Unified ELO 1446 ± 13, rank #1015 of 1776 rated models, from 68 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LatamBoard - Spanish COPA | 85 | Score (%) | 97 |
| LatamBoard - Spanish WNLI | 80.28 | Score (%) | 95.5 |
| LatamBoard - Spanish PAWS | 63.8 | Score (%) | 93.9 |
| Open CoT - LSAT Reading Comprehension | 22.3 | CoT Gain (%) | 93.1 |
| Open CoT - LogiQA 2 | 15.01 | CoT Gain (%) | 92.7 |
| LatamBoard - Spanish OpenBookQA | 38.8 | Score (%) | 90.9 |
| Open Chinese LLM - TruthfulQA MC | 61.29 | Accuracy (%) | 88.7 |
| LatamBoard - FaQuAD NLI | 78.57 | Score (%) | 87.9 |
| LatamBoard - Portuguese Score | 87.15 | Score (%) | 87.9 |
| LatamBoard - Spanish TeleIA | 71.43 | Score (%) | 86.4 |
| Open CoT Leaderboard | 12.85 | Average CoT Gain (%) | 86.3 |
| LatamBoard - Spanish Score | 62.61 | Score (%) | 84.8 |
Interactive version: theaggregate.ai/model?slug=solar-10-7b-instruct-v1-0 · How the rankings work · Data refreshed daily, snapshot 2026-07-22.