Qwen · open-weights

Qwen2-72B

Qwen2-72B is an open-weights language model from Qwen, released June 7, 2024. It has 73B parameters (dense). The context window is 128K tokens. On the Epoch Capabilities Index it scores 122.3, rank 179 of 256 rated models.

Check it against your own PC →

Lab

Qwen

Released

June 7, 2024

Weights

Open tongyi-qianwen

Parameters

73B dense

Context

128K

Price / M tokens

–

Epoch ECI

122.3 #179

Can you run Qwen2-72B locally?

The smallest common GPU it runs on is 24 GB of VRAM (with 32 GB of RAM): IQ2_XXS, split GPU / CPU, roughly 12 tokens per second. Estimates assume 32 GB of DDR5 RAM, a 16K context and llama.cpp-style GGUF quantization. Enter your exact hardware in the live app.

GPU memory	Verdict	Best quant	VRAM	RAM	tok/s
6 GB VRAM	Too large	–	–	–	–
8 GB VRAM	Very slow	IQ2_XXS (heavy loss)	7.2 GB	17.5 GB	~2
12 GB VRAM	Very slow	IQ2_XXS (heavy loss)	11.2 GB	13.5 GB	~2
16 GB VRAM	Very slow	IQ2_XXS (heavy loss)	15.2 GB	9.5 GB	~3
24 GB VRAM	Runs, slowly	IQ2_XXS (heavy loss)	23.2 GB	1.5 GB	~12
32 GB VRAM	Runs on GPU	IQ2_M (heavy loss)	30.5 GB	–	~42

Ready-made quantizations on Hugging Face: bartowski/dolphin-2.9.2-qwen2-72b-GGUF · mradermacher/Qwen2-72B-i1-GGUF · mradermacher/Qwen2-72B-GGUF · featherless-ai-quants/Qwen-Qwen2-72B-GGUF · BlouseJury/cognitivecomputations_dolphin-2.9.2-qwen2-72b-Q6_K-gguf · bigstorm/dolphin-2.9.2-qwen2-72b-7.0bpw-8hb-exl2

Strengths by category

ModelTape's 0–100 fit on Epoch AI benchmark results.

Category	Score	Based on
Knowledge	30.0	GPQA diamond, MMLU

Benchmark results

Benchmark (Epoch AI)	Score
MMLU	76.5%
MATH level 5	39.1%
GPQA Diamond	21.9%
The Agent Company	1.1%

Links

Hugging Face: Qwen/Qwen2-72B · Epoch AI benchmarks

Models with a similar score

Llama 3.3 70B · GPT-4 Turbo (Apr 2024) · Claude 3 Opus · Gemini 1.5 Pro (May 2024) · Claude 3.5 Haiku · GPT-4 Turbo (Nov 2023) · Llama 3.1-70B · GPT-4o mini