GritLM-8x7B-KTO Benchmark Results

GritLM-8x7B-KTO — benchmark results

Contextual AI's KTO-tuned GritLM 8x7B, a Mixtral-based GRIT model that unifies text generation and embedding in one checkpoint. Provider: Other. Released 2024-04-17. Access: Open.

Unified ELO 1432 ± 8, rank #1081 of 1776 rated models, from 14 benchmark results.

Strongest benchmark results

Benchmark	Score	Metric	Percentile
Open LLM Leaderboard - BBH	40.83	Score	79.8
Open Chinese LLM - WinoGrande	65.59	Accuracy (%)	73.3
Open LLM Leaderboard - IFEval	57.14	Score	69
Open Chinese LLM - ARC Challenge	54.35	Accuracy (%)	65.3
Open Chinese LLM Leaderboard	58.43	Average Score (%)	64.7
Open LLM Leaderboard - MuSR	11.67	Score	62.4
Open Chinese LLM - HellaSwag	61.31	Accuracy (%)	62
Open Chinese LLM - C-Eval Semantic	72.97	Accuracy (%)	60.5
Open Chinese LLM - CMMLU	56.28	Accuracy (%)	60.5
Open LLM Leaderboard - MMLU-Pro	29.42	Score	56.8
Open LLM Leaderboard - MATH Level 5	12.24	Score	55
Open LLM Leaderboard - GPQA	6.15	Score	51.6

Interactive version: theaggregate.ai/model?slug=gritlm-8x7b-kto · How the rankings work · Data refreshed daily, snapshot 2026-07-22.