Meta · open-weights

Llama 3.1-70B

Llama 3.1-70B is an open-weights language model from Meta, released July 23, 2024. It has 71B parameters (dense). On the Epoch Capabilities Index it scores 124.7, rank 176 of 256 rated models.

Check it against your own PC →

Lab

Can you run Llama 3.1-70B locally?

The smallest common GPU it runs on is 24 GB of VRAM (with 32 GB of RAM): IQ2_XXS, fully on GPU, roughly 30 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	Very slow	IQ2_XXS (heavy loss)	5.2 GB	17.9 GB	~2
8 GB VRAM	Very slow	IQ2_XXS (heavy loss)	7.2 GB	15.9 GB	~2
12 GB VRAM	Very slow	IQ2_XXS (heavy loss)	11.2 GB	11.9 GB	~2
16 GB VRAM	Very slow	IQ2_XXS (heavy loss)	15.2 GB	7.9 GB	~3
24 GB VRAM	Runs on GPU	IQ2_XXS (heavy loss)	23.1 GB	–	~30
32 GB VRAM	Runs on GPU	IQ2_M (heavy loss)	28.8 GB	–	~43

Ready-made quantizations on Hugging Face: NousResearch/Hermes-3-Llama-3.1-70B-GGUF · mradermacher/Meta-Llama-3.1-70B-i1-GGUF · ThomasBaruzier/Llama-3.3-70B-Instruct-GGUF · Mungert/Llama-3.1-70B-Instruct-GGUF · Mungert/Llama-3.3-70B-Instruct-GGUF · casperhansen/llama-3.3-70b-instruct-awq

Strengths by category

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

Category	Score	Based on
Math	11.4	MATH level 5, OTIS Mock AIME 2024-2025
Agent	17.1	Balrog, The Agent Company
Knowledge	29.4	GPQA diamond, MMLU

Benchmark results

Benchmark (Epoch AI)	Score
MMLU	73.5%
MATH level 5	36.7%
Balrog	27.9%
GPQA Diamond	26.7%
The Agent Company	6.9%
OTIS Mock AIME	3.5%

Links

Hugging Face: meta-llama/Llama-3.1-70B · Epoch AI benchmarks

Models with a similar score

GPT-4o (May 2024) · Llama 3.1-405B · Llama 4 Scout · Gemma 3 27B · Qwen2.5-72B · GPT-4o (Aug 2024) · Gemini 1.5 Flash (Sep 2024) · GPT-4.1 nano