Llama 3.3 70B

Name: Llama 3.3 70B
Price: 0.155 USD
Author: Meta

Meta · Llama 3 · released 2024-12-06 · data checked 2026-11-01 Needs verification

Llama 3.3 70B is a 70.6B-parameter dense model from Meta (United States), released 2024-12-06 under the Llama 3.3 Community license. It scores 40.8 on the Small LM Index, #6 of 6 large models. It needs about 47.4 GB of memory at 4-bit with an 8K-token context and costs $0.15 per million tokens through the cheapest API provider.

Small LM Index

40.8

#6 of 6 scored large models

Likely 30.9–51.1 · from 2 of 6 core benchmarks

Parameters

70.6B

Dense

Price per 1M tokens

$0.15

$0.10 in / $0.32 out · blended 3:1

Context window

128K

tokens

Benchmarks

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

MMLU-Pro Lab-reported Knowledge68.9

Class best: Qwen3-Next 80B-A3B Thinking 82.7

GPQA Diamond Lab-reported Knowledge50.5

Class best: Nemotron 3.5 Bolt 74B-A9B 88.8

Not reported: estimated

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

Humanity's Last Exam≈ 4

AIME≈ 54

LiveCodeBench≈ 45

SWE-bench Verified≈ 21

Specifications

Lab: Meta (US · United States)
Architecture: Dense transformer
Total parameters: 70.6B
Active parameters: All (dense)
Size class: Large
Context window: 128K tokens
Input modalities: text
Reasoning: Instruct (no thinking)
License: Llama 3.3 Community
Openness: Open weights
Memory to run: ~47.4 GB at 4-bit, 8K context (more sizes)
Hugging Face: 363.3K downloads (30 days) · 3.2K likes
API providers: 8 on OpenRouter
API price: $0.10 in / $0.32 out per 1M (OpenRouter, 2026-11-03)

Weights on Hugging Face ↗Try on OpenRouter ↗

Compare with

Llama 3.3 70B vs Nemotron 3.5 Bolt 74B-A9B (74B, index 73.3)
Llama 3.3 70B vs Kolibri-1 (78.1B, index 68.3)
Llama 3.3 70B vs Qwen3-Next 80B-A3B Thinking (80B, index 63.3)
Llama 3.3 70B vs K2-V2 Instruct (70B, index 57.9)
Llama 3.3 70B vs DeepSeek-R1-Distill-Llama-70B (70.6B, index 50.1)

Sources

https://huggingface.co/meta-llama/Llama-3.3-70B-Instruct

Something wrong here? Report it.

Run it locally

Uses unsloth/Llama-3.3-70B-Instruct-GGUF (Q4_K_M). Swap the tag for another size: BF16, F16, IQ4_NL, IQ4_XS, Q2_K, Q2_K_L, Q3_K_M, Q3_K_S….

Ollama

ollama run hf.co/unsloth/Llama-3.3-70B-Instruct-GGUF:Q4_K_M

llama.cpp

llama-server -hf unsloth/Llama-3.3-70B-Instruct-GGUF:Q4_K_M

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

MLX (Mac)

pip install -U mlx-lm
mlx_lm.chat --model mlx-community/Llama-3.3-70B-Instruct-4bit

vLLM (GPU server)

vllm serve meta-llama/Llama-3.3-70B-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	144 GB	154 GB	185 GB
8-bit Q8_0	78 GB	88 GB	119 GB
4-bit Q4_K_M	46 GB	55 GB	87 GB

Check it against your GPU, Mac or phone →