Llama 3.1 8B

Name: Llama 3.1 8B
Price: 0.025 USD
Author: Meta

Meta · Llama 3 · released 2024-07-23 · data checked 2026-11-01 Needs verification

Llama 3.1 8B is a 8B-parameter dense model from Meta (United States), released 2024-07-23 under the Llama 3.1 Community license. It scores 21.3 on the Small LM Index, #23 of 26 small models. It needs about 6.6 GB of memory at 4-bit with an 8K-token context and costs $0.025 per million tokens through the cheapest API provider.

Small LM Index

21.3

#23 of 26 scored small models

Likely 15.3–29.2 · from 2 of 6 core benchmarks

Parameters

Dense

Price per 1M tokens

$0.025

$0.020 in / $0.040 out · blended 3:1

Context window

128K

tokens

Benchmarks

Each bar shows this model's score. The tick marks the best small model on that benchmark. Sources: 2 reported by Meta. How we label sources.

MMLU-Pro Lab-reported Knowledge48.3

Class best: Qwen3.5 9B 82.5

GPQA Diamond Lab-reported Knowledge30.4

Class best: Qwen3.5 9B 81.7

Not reported: estimated

What the index model expects from this model's other scores. Not a measurement.

Humanity's Last Exam≈ 1

AIME≈ 18

LiveCodeBench≈ 21

SWE-bench Verified≈ 7

Specifications

Lab: Meta (US · United States)
Architecture: Dense transformer
Total parameters: 8B
Active parameters: All (dense)
Size class: Small
Context window: 128K tokens
Input modalities: text
Reasoning: Instruct (no thinking)
License: Llama 3.1 Community
Openness: Open weights
Memory to run: ~6.6 GB at 4-bit, 8K context (more sizes)
Hugging Face: 4.5M downloads (30 days) · 8.8K likes
API providers: 4 on OpenRouter
API price: $0.020 in / $0.040 out per 1M (OpenRouter, 2026-11-03)

Weights on Hugging Face ↗Try on OpenRouter ↗

Compare with

Llama 3.1 8B vs Granite 4.2 8B (8B, index 63.5)
Llama 3.1 8B vs Ministral 3 8B (8B, index 55.2)
Llama 3.1 8B vs Gemma 4 E4B (8B, index 38.2)
Llama 3.1 8B vs Ling 3.0 Tiny (7.9B, index 51.9)

Sources

https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct

Something wrong here? Report it.

Run it locally

Uses ggml-org/Meta-Llama-3.1-8B-Instruct-Q4_0-GGUF (Q4_0). Swap the tag for another size: Q4_0.

Ollama

ollama run hf.co/ggml-org/Meta-Llama-3.1-8B-Instruct-Q4_0-GGUF:Q4_0

llama.cpp

llama-server -hf ggml-org/Meta-Llama-3.1-8B-Instruct-Q4_0-GGUF:Q4_0

Then open http://localhost:8080 for a chat window.

MLX (Mac)

pip install -U mlx-lm
mlx_lm.chat --model mlx-community/Llama-3.1-8B-Instruct-4bit

vLLM (GPU server)

vllm serve meta-llama/Llama-3.1-8B-Instruct

Full-precision weights, for a server GPU.

Open in LM Studio

Memory needed: weights, context cache and runtime buffers.

Quantization	4K context	32K context	128K context
16-bit F16	17 GB	21 GB	33 GB
8-bit Q8_0	9.7 GB	13 GB	26 GB
4-bit Q4_K_M	6.0 GB	9.8 GB	22 GB

Check it against your GPU, Mac or phone →