DeepSeek-R1-Distill-Llama-70B

Name: DeepSeek-R1-Distill-Llama-70B
Author: DeepSeek

DeepSeek · DeepSeek-R1 distill · released 2025-01-20 · data checked 2026-11-01 Needs verification

DeepSeek-R1-Distill-Llama-70B is a 70.6B-parameter dense model from DeepSeek (China), released 2025-01-20 under the MIT license. It scores 50.1 on the Small LM Index, #5 of 6 large models. It needs about 47.4 GB of memory at 4-bit with an 8K-token context.

Small LM Index

50.1

#5 of 6 scored large models

Likely 45.3–54.8 · from 3 of 6 core benchmarks

Parameters

70.6B

Dense

Price per 1M tokens

—

No hosted API on OpenRouter; self-host

Context window

128K

tokens

Benchmarks

Each bar shows this model's score. The tick marks the best large model on that benchmark. Sources: 3 reported by DeepSeek. How we label sources.

GPQA Diamond Lab-reported Knowledge65.2

Class best: Nemotron 3.5 Bolt 74B-A9B 88.8

AIME Lab-reported Math · AIME 202470.0

Class best: Kolibri-1 96.0

LiveCodeBench Lab-reported Coding57.5

Class best: Kolibri-1 85.9

Not reported: estimated

What the index model expects from this model's other scores. Not a measurement.

MMLU-Pro≈ 72

Humanity's Last Exam≈ 6

SWE-bench Verified≈ 31

Specifications

Lab: DeepSeek (CN · China)
Architecture: Dense transformer
Total parameters: 70.6B
Active parameters: All (dense)
Size class: Large
Context window: 128K tokens
Input modalities: text
Reasoning: Reasoning (thinking)
License: MIT
Openness: Open weights
Memory to run: ~47.4 GB at 4-bit, 8K context (more sizes)
Hugging Face: 30.5K downloads (30 days) · 823 likes

Weights on Hugging Face ↗

Compare with

DeepSeek-R1-Distill-Llama-70B vs Nemotron 3.5 Bolt 74B-A9B (74B, index 73.3)
DeepSeek-R1-Distill-Llama-70B vs Kolibri-1 (78.1B, index 68.3)
DeepSeek-R1-Distill-Llama-70B vs Qwen3-Next 80B-A3B Thinking (80B, index 63.3)
DeepSeek-R1-Distill-Llama-70B vs K2-V2 Instruct (70B, index 57.9)
DeepSeek-R1-Distill-Llama-70B vs Llama 3.3 70B (70.6B, index 40.8)

Sources

https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Llama-70B

Something wrong here? Report it.

Run it locally

Uses unsloth/DeepSeek-R1-Distill-Llama-70B-GGUF (Q4_K_M). Swap the tag for another size: BF16, F16, IQ4_NL, IQ4_XS, Q2_K, Q2_K_L, Q3_K_M, Q3_K_S….

Ollama

ollama run hf.co/unsloth/DeepSeek-R1-Distill-Llama-70B-GGUF:Q4_K_M

llama.cpp

llama-server -hf unsloth/DeepSeek-R1-Distill-Llama-70B-GGUF:Q4_K_M

Then open http://localhost:8080 for a chat window.

MLX (Mac)

pip install -U mlx-lm
mlx_lm.chat --model mlx-community/DeepSeek-R1-Distill-Llama-70B-4bit

vLLM (GPU server)

vllm serve deepseek-ai/DeepSeek-R1-Distill-Llama-70B

Full-precision weights, for a server GPU.

Open in LM Studio

Memory needed: weights, context cache and runtime buffers.

Quantization	4K context	32K context	128K context
16-bit F16	144 GB	154 GB	185 GB
8-bit Q8_0	78 GB	88 GB	119 GB
4-bit Q4_K_M	46 GB	55 GB	87 GB

Check it against your GPU, Mac or phone →