dolphin-2.9.1-llama-3-70B — benchmark results

Eric Hartford's Cognitive Computations uncensored Dolphin fine-tune of Llama 3 70B with agentic and function-calling abilities. Provider: Cognitive Computations. Released 2024-05-22. Access: Open.

Unified ELO 1490 ± 27, rank #841 of 1776 rated models, from 14 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Open LLM Leaderboard - MuSR23.7Score98.5
Open Chinese LLM - GSM8K70.96Accuracy (%)97.6
Open Chinese LLM - CMMLU72.34Accuracy (%)96.1
Open Chinese LLM Leaderboard68.63Average Score (%)92.3
Open Chinese LLM - ARC Challenge60.49Accuracy (%)88.9
Open Chinese LLM - HellaSwag67.85Accuracy (%)85.2
Open Chinese LLM - WinoGrande67.72Accuracy (%)84.9
Open Chinese LLM - C-Eval Semantic87.12Accuracy (%)83.7
Open LLM Leaderboard - MMLU-Pro34.78Score75.2
Open LLM Leaderboard - MATH Level 518.2Score67.8
Open LLM Leaderboard - GPQA7.83Score66.2
Open LLM Leaderboard - BBH31.1Score55.6

Interactive version: theaggregate.ai/model?slug=dolphin-2-9-1-llama-3-70b · How the rankings work · Data refreshed daily, snapshot 2026-07-22.