128K tokens
Context window
Llama 3.3 70B
Private data, offline work, fine-tuning
8K tokens
Max output
Local
Input per 1M
Local
Output per 1M
late 2023
Knowledge cutoff
text
Modalities
Last verified 20 May 2026 Report a correction
What it tends to refuse
Base weights refuse very little; behaviour depends on the instruct tuning you run
At 4-bit quantisation this needs roughly 42 GB plus room for the context, so two 24 GB cards or one 48 GB card. Follows multi-part instructions less reliably than hosted models.
Standing rules for this model go in Modelfile, not in every message.
See the config files.
Written for Llama
All 6- CheatsheetLlama
Ollama commands worth knowing
The subset of the CLI you actually use, plus the two flags that fix the problems everybody hits.
Tested on Llama 3.3 70B · May 2026
- Local setupLlama
Running your first local model
Ollama on a normal laptop, from install to a model answering in about ten minutes, plus how to work out which size actually fits before you download…
Tested on Llama 3.2 8B · May 2026
LM Studio
A desktop app for downloading, running and comparing local models without a terminal. Best tool for deciding which quantisation is good enough before…
Tested on Llama 3.2 8B · May 2026
Ollama
The simplest way to get a local model answering on your own machine. One command to install, one to pull a model, and an API on localhost that most…
Tested on Llama 3.3 70B · May 2026
- GuideLlamaMistral
Running a model on your own machine
What local models are genuinely good for, what hardware you need, and what to change in prompts that came from a hosted model.
- ConfigLlamaMistral
Ollama Modelfile for a local code assistant
A Modelfile with a system prompt and parameters tuned for short, correct code answers on a local model.