Qwen

Qwen2.5-72B-Instruct

chatmodelsQwen2ForCausalLMbfloat16

Install and run this model locally using llmpm, the open-source LLM package manager.

Install
llmpm install Qwen/Qwen2.5-72B-Instruct
Run
llmpm run Qwen/Qwen2.5-72B-Instruct
Average Score (0–100)
48.0%
Weighted average of normalized scores from all benchmarks. Each benchmark is normalized to a 0–100 scale, then averaged together.

BENCHMARK SCORES

IFEval86.4%

Instruction-Following Evaluation. Tests the model's ability to follow explicit formatting instructions (instruction following, formatting, generation). Scored by strict format accuracy.

BBH61.9%

Big Bench Hard. A collection of challenging tasks across language understanding, mathematical reasoning, and common sense knowledge. Scored by accuracy on multiple-choice questions.

MATH Lvl 559.8%

Mathematics Aptitude Test of Heuristics, Level 5. High school competition problems covering complex algebra, geometry, and advanced calculus. Scored by exact match.

GPQA16.7%

Graduate-Level Google-Proof Q&A. PhD-level multiple-choice questions in chemistry, biology, and physics. Scored by accuracy.

MuSR11.7%

Multistep Soft Reasoning. Tests reasoning and understanding over long texts, including language understanding, reasoning capabilities, and long-context reasoning. Scored by accuracy.

MMLU-Pro51.4%

Massive Multitask Language Understanding – Professional. Expert-reviewed multiple-choice questions across medicine, law, engineering, and mathematics. Scored by accuracy.

MODEL INFO

Architecture
Qwen2ForCausalLM
Precision
bfloat16
Type
chatmodels
Weight Type
Original
Parameters
72.7B
Chat Template
Yes

METADATA

Upload Date
2024-09-16
Submission Date
2024-10-16
License
other
Base Model
Qwen/Qwen2.5-72B
HF Hearts
777
CO₂ Cost (kg)
47.65