Model catalog

Llama 3.3 70B Instruct

Meta's 70B text-only instruction model, extending the catalog into the high-memory class.

Check this model on your hardware See what your PC can run

Model overview

Family: Llama 3.3
Provider: Meta
Architecture: Llama
Parameters: 70B
Formats: gguf, safetensors
Runtimes: llama.cpp
License: Llama 3.3 Community License
Maximum context: 131,072 tokens

Available quantizations

Lower-bit quantizations generally use less memory. The compatibility calculator uses these candidates and its transparent approximate memory assumptions; it does not make a performance guarantee.

Q4_K_M

A commonly available lower-memory GGUF variant; the listed memory input is an estimate, not the repository file size.

Bits per weight: 4.5
Source: bartowski/Llama-3.3-70B-Instruct-GGUF GGUF repository

Q8_0

A higher-memory GGUF variant where listed by the conversion repository; the listed memory input remains approximate.

Bits per weight: 8
Source: bartowski/Llama-3.3-70B-Instruct-GGUF GGUF repository

Compatibility notes

Weight estimates are approximate and include a runtime overhead. Real requirements vary with context length, runtime, drivers, and other system use. A model page cannot determine compatibility without your hardware profile; use the calculator for that assessment.

Meta Llama 3.3 70B Instruct model card; verified value, last verified 2026-09-24. Meta documents the text-only 70B variant, 128K context, and Llama 3.3 Community License. The GGUF candidates are from bartowski's separate community conversion and remain subject to the model's license. View source