Skip to content
local-ai

Llama 3.3 70B Instruct

Meta · Text generation · 70B · 131k context · Released 6 December 2024

Permitted with conditions Open weights Runs on CPU Apple Silicon

Meta's 70-billion-parameter instruction-tuned model, delivering performance close to their much larger 405B model at a fraction of the hardware cost. A strong general-purpose choice if you have the VRAM for it.

Strengths

  • Strong general reasoning and instruction following
  • Long 128k context window
  • Excellent ecosystem support across every major inference engine

Weaknesses

  • Needs serious hardware, not a laptop model at usable quality
  • Weaker at code than dedicated coding models of similar size
  • Licence carries conditions that matter at scale

Hardware requirements

QuantisationApprox. VRAMNotes
Q4_K_M~43GBGood balance, the common choice for 48GB cards
Q5_K_M~50GBNoticeably better than Q4, needs more headroom
Q8_0~75GBNear-lossless, but you need multiple GPUs
FP16~141GBFull precision, server territory

Also runs on CPU (slower). Optimised builds available for Apple Silicon.

What you'd need to run this

Roughly what a machine to run this would need, at up to three levels of quality. Memory is the deciding factor.

Minimum to run it

Q4_K_M · ~43GB needed

64GB of unified memory

Mac mini M4 Pro

or an 80GB-class data-centre card, which is usually rented by the hour, NVIDIA A100 80GB .

Unified memory is shared with the model, so it is already counted above.

around £1,100–£2,000

See the Affordable Mac build build What else 64GB runs →

For good quality

Q5_K_M · ~50GB needed

64GB of unified memory

Mac mini M4 Pro

or an 80GB-class data-centre card, which is usually rented by the hour, NVIDIA A100 80GB .

Unified memory is shared with the model, so it is already counted above.

around £1,100–£2,000

See the Affordable Mac build build What else 64GB runs →

Best quality

FP16 · ~141GB needed

A multi-GPU server, roughly 3× 80GB-class GPUs

Server memory in the hundreds of gigabytes or more, plus fast interconnect.

tens of thousands of pounds, or rented by the hour

Usually rented — see the cost calculator

Licence

Llama 3.3 Community License — read the licence

Benchmarks

BenchmarkScoreSourceAs of
MMLU86.0 Meta model card December 2024
Artificial Analysis Intelligence Index v4.1.19 Artificial Analysis (independent, figure marked estimated) August 2026

How it compares

How this model’s reported scores sit against other models we cover, on the same benchmarks. This model is highlighted.

Artificial Analysis Intelligence Index v4.1.1

higher is better

MMLU

higher is better

Llama 3.3 70B Instruct: common questions

What hardware do I need to run Llama 3.3 70B Instruct?
At its most compressed (Q4_K_M) it needs roughly 43GB of VRAM, and about 50GB for good quality. VRAM figures are approximate and depend on context length and settings.
Is Llama 3.3 70B Instruct free for commercial use?
Commercial use is permitted, but with conditions. Commercial use is permitted, but organisations with more than 700 million monthly active users must request a separate licence from Meta. Read the licence before relying on it at scale.
Can I run Llama 3.3 70B Instruct on Apple Silicon?
Yes. Llama 3.3 70B Instruct has builds optimised for Apple Silicon, through MLX or GGUF on a Mac.
Does Llama 3.3 70B Instruct run on CPU?
Yes, Llama 3.3 70B Instruct can run on the CPU, though generation is slower than on a GPU.
What is Llama 3.3 70B Instruct's context window?
Llama 3.3 70B Instruct has a context window of 131,072 tokens, about 131k.

Availability

Where to get quantised weights

Some of the best quantised weights are made by the community, not the model’s authors. Look this model up on these providers:

  • Unsloth GGUF (Dynamic 2.0, imatrix)

    Dynamic quants hold quality well at Q4, which is what a 24GB card needs.

  • Bartowski GGUF Q2-Q8 (imatrix)

    A wide, reliable range of imatrix GGUF quants, typically Q2 through Q8.

  • mradermacher

    Very broad coverage of models, in both static and imatrix GGUF.

  • MLX community MLX 4-bit and 8-bit

    For Apple Silicon, with 4-bit suiting 48GB or more of unified memory.

Recommended for

  • General-purpose assistant on a 48GB+ setup
  • Long-document analysis
  • Mac Studio users with 64GB+ unified memory

Run it with

Related guides

Glossary

Our coverage

Catalogue entry last verified 30 July 2026. Specifications change; verify anything you are about to spend money on.