# Llama 3.3 70B Instruct (Q4_K_M)

> Meta · 70.6B · q4_k_m

Canonical: https://gpuninja.com/models/llama-3-3-70b-instruct-q4-k-m
Data as of: 2026-10-11T15:48:18Z
Path: Home › AI models
- Specs verified: 11 Oct 2026, 15:48 UTC

## Specifications

- Publisher: Meta (Verified · source github.com · retrieved 11 Oct 2026, 15:42 UTC · checked 11 Oct 2026, 15:48 UTC)
- Parameters: 70.6B (Verified · source huggingface.co · retrieved 11 Oct 2026, 15:40 UTC · checked 11 Oct 2026, 15:48 UTC)
- Quantization: q4_k_m (Verified · source huggingface.co · retrieved 11 Oct 2026, 15:40 UTC · checked 11 Oct 2026, 15:48 UTC)
- Minimum VRAM: 42.5 GB (Unverified · source huggingface.co · retrieved 11 Oct 2026, 15:40 UTC · checked 11 Oct 2026, 15:48 UTC)
- Context length: 128K tokens (Verified · source huggingface.co · retrieved 11 Oct 2026, 15:40 UTC · checked 11 Oct 2026, 15:48 UTC)
- Modality: text (Verified · source github.com · retrieved 11 Oct 2026, 15:42 UTC · checked 11 Oct 2026, 15:48 UTC)
- License: Llama 3.3 Community License (Verified · source github.com · retrieved 11 Oct 2026, 15:42 UTC · checked 11 Oct 2026, 15:48 UTC)

## Graphics cards that can run it

Board, Chip, VRAM, TDP, Length, Slots, Price, Stock, VRAM headroom

No graphics cards recorded that can run this model.

## Benchmarks

No benchmark results recorded.

## Sources

- https://github.com/meta-llama/llama-models/blob/main/models/llama3_3/MODEL_CARD.md — Publisher, Modality, License; last verified 11 Oct 2026, 15:48 UTC
- https://huggingface.co/unsloth/Llama-3.3-70B-Instruct — Parameters; last verified 11 Oct 2026, 15:48 UTC
- https://huggingface.co/bartowski/Llama-3.3-70B-Instruct-GGUF/blob/main/Llama-3.3-70B-Instruct-Q4_K_M.gguf — Quantization, Minimum VRAM; last verified 11 Oct 2026, 15:48 UTC
- https://huggingface.co/unsloth/Llama-3.3-70B-Instruct/blob/main/config.json — Context length; last verified 11 Oct 2026, 15:48 UTC