Reasoning & Math Models · Fast-moving · Advanced

DeepSeek R1 Distill Llama 70B

Also known as: R1-Distill-70B, DeepSeek Llama 70B

A 70B parameter reasoning model distilled from DeepSeek R1 into Llama 3.3 70B Instruct architecture.

What DeepSeek R1 Distill Llama 70B is

Delivers near-frontier reasoning performance on math and coding benchmarks while fitting on dual GPU server nodes.

How it works

Trained on high-quality R1 reasoning datasets combined with curated supervised fine-tuning corpora.

Why it matters

Ideal for enterprise deployments requiring open weights, high reasoning accuracy, and standard Llama ecosystem compatibility.

Common uses