DeepSeek-R1

DeepSeek-R1-Distill-Llama-8B

The DeepSeek-R1 card names this distill 8B, based on Llama-3.1-8B. The card does not publish a separate exact parameter total, so this page uses the 8B size class.

MIT, plus the Llama 3.1 Community License · 8B · context 131,072

config.json max_position_embeddings is 131072, matching the Llama 3.1 128k context.

From the model card

Strengths

  • The DeepSeek-R1 card says these dense models are distilled from DeepSeek-R1's reasoning traces.
  • The 8B distill is based on Llama-3.1-8B.

Weaknesses

  • Two licenses apply: MIT on DeepSeek's files, and the Llama 3.1 Community License on the base.
  • Reasoning traces spend context and time.
  • No Frankenstein row yet. Text only.

Sources: DeepSeek-R1 model card · Distill model card · Llama 3.1 license · config.json

Memory at 8,192 tokens

Computed, not measured. Weights use llama.cpp’s whole-file averages (Q4_K_M = 4.8944 bits/weight, Q8_0 = 8.5008) on the total parameter count. KV is fp16 and does not depend on the quant. The formula, including the sliding-window and latent-KV cases, is on the methodology page.

Computed memory for DeepSeek-R1-Distill-Llama-8B at 8192 context
Q4_K_MQ8_0
Weights4.56 GiB7.92 GiB
KV at 8,1921 GiB1 GiB
Total5.56 GiB Computed8.92 GiB Computed

32 layers, 8 KV heads, head dim 128. Every layer stores the full context.

Tokens per second

Not yet measured. This lab has no Frankenstein row for DeepSeek-R1-Distill-Llama-8B, so there is no estimated tok/s either.

Hardware tiers

Minimum is the smallest RAM that holds Q4_K_M plus 8k KV. Suggested does not name a second machine or a speed this lab has not run. Performance means a CUDA-class card with 12 GB or more of GDDR6 or GDDR6X and the weights fully on the card. The Tesla P40 is 24 GB of GDDR5, so a P40 speed is not the performance-tier number.

TierExample hardwareQuantFittok/s
MinimumCPU-only, with at least 7.56 GiB of system RAM for the Q4_K_M weights, 8k KV, and a 2 GiB planning allowance.Q4_K_MCPU-only, if the RAM is thereCPU: not yet measured
SuggestedA machine you are willing to wait on, or a GPU that can hold the quant. This lab has not measured a separate “comfortable” machine, so this row does not invent one.Q4_K_M until a Q8 run existsWhichever of the machines above actually holds the weightsnot yet measured
PerformanceA CUDA-class card with 12 GB or more of GDDR6 or GDDR6X, weights fully on the card.Q4_K_MFull VRAM on a 12 GB or larger GDDR6-class cardnot yet measured

Where it sits

Best-for is a judgment from the card’s stated jobs and the model’s size. It is not a speed ranking. The matching build is Cheap 8B chat. Score it against a machine you have on What can my rig run?