LLMpopular

Llama 3.3 70B

Meta · 70B · 128k ctx

Meta's flagship open-weights model. Pairs well with QLoRA when you want to ship fast on a single node, or full FT when you have the budget and >100M tokens of high-quality data.

Train this model →Read the recipe

Quickstart

Launch a Llama 3.3 70B fine-tune

~ fourbit

$ fourbit train \
  --base llama-3-3-70b \
  --data s3://acme/dataset.jsonl \
  --recipe lora

More LLMs

Qwen3 72B
Strong multilingual generalist with long context. Excellent for non-English and code tasks.
72B
Llama 3.1 8B
Workhorse 8B. Cheap to tune, fast to serve, ships on one GPU.
8B
Qwen3 8B
8B with long context. Strong multilingual and code performance per dollar.
8B