Fourbitfourbit
Request access
LLM

Llama 3.1 8B

Meta · 8B · 128k ctx

The default starting point. Most domain adaptations should start here — if 8B can't reach your eval bar, then move up to 70B-class.

Train this model →Read the recipe
Quickstart

Launch a Llama 3.1 8B fine-tune

~ fourbit
$ fourbit train \
  --base llama-3-1-8b \
  --data s3://acme/dataset.jsonl \
  --recipe lora
More LLMs