Active Parameters
671B
Context Length
131K
Modality
Text
Architecture
Mixture of Experts
KV Cache Rate
69 KB/tok
License
MIT License
Release Date
27 Dec 2024
Knowledge Cutoff
Jul 2024
API Pricing (per 1M)
Input: $0.50 · Output: $2.15
Precision
Context
Workload Memory: 1325.07 GB
FP16 · 8K context
Base Weights: 1318.75 GB
KV Cache: 3.81 GB
Runtime & Graphs: 2.50 GB
16x NVIDIA H200 141GB
1325.1 / 2030.4 GB usable · TP 8 · PP 2
Determine requirements for your specific workload:
Upgrade to Pro
Checkpoint format
FP16
CUDA
>= 11.8
PyTorch
>= 2.1
Transformers
>= 4.46.3
| Architecture | Precision | Attention kernel |
|---|---|---|
Blackwell, Hopper | Native FP16 | FlashMLA |
Ada Lovelace, Ampere | Native FP16 | FlashAttention-2 |
Additional required packages
Install alongside standard runtime libraries:
pip install flash_attn
Rank
#135
| Benchmark | Score | Rank |
|---|---|---|
Professional Knowledge | 0.85 | 13 |
StackUnseen | 0.524 | 21 |
Graduate-Level QA | 0.81 | 62 |
General Text | 1422 | 102 |
Agentic Index | 0.01 | 113 |
Coding Index | 0.25 | 124 |
Intelligence Index | 0.13 | 233 |
StackEval Archived | 0.956 | 5 |
QA Assistant Archived | 0.964 | 5 |
Coding Archived | 0.71 | 7 |
Summarization Archived | 0.774 | 14 |
Overall Rank
#135
Coding Rank
#119
DeepSeek-R1 is DeepSeek's flagship 671B Mixture-of-Experts reasoning model engineered with large-scale reinforcement learning for complex problem-solving. It generates explicit multi-step chains of thought for state-of-the-art math, coding, and scientific deduction.
Attention
Attention Structure
Multi-Layer Attention
Attention Heads
128
Key-Value Heads
128
Position Embedding
ROPE
RoPE Theta
10,000
Normalization
RMS Normalization
Activation Function
Swish
Dimensions
Base Precision
FP16
Hidden Dimension Size
2,048
Number of Layers
61
FFN Intermediate Size (Dense)
2,048
Multi-Token Prediction Heads
1
Tokenizer
Vocabulary Size
129,280
Mixture of Experts
Total Expert Parameters
37.0B
Number of Experts
64
Active Experts
6
Shared Experts
1
FFN Intermediate Size (per Expert)
2,048
Dense Layers Before MoE
3
DeepSeek-R1 is a model family developed for logical reasoning tasks. It incorporates a Mixture-of-Experts architecture for computational efficiency and scalability. The family utilizes Multi-Head Latent Attention and employs reinforcement learning in its training, with some variants integrating cold-start data.