Skip to content | AI model | Parameters | Active parameters | Layers | KV heads | Context |
| DeepSeek deepseek-coder-1.3b-instruct | 1.35 B | – | 24 layers | 16 heads | 16,384 tokens |
|---|
| DeepSeek deepseek-coder-6.7b-base | 6.74 B | – | 32 layers | 32 heads | 16,384 tokens |
|---|
| DeepSeek deepseek-coder-6.7b-instruct | 6.74 B | – | 32 layers | 32 heads | 16,384 tokens |
|---|
| DeepSeek deepseek-coder-7b-instruct-v1.5 | 6.91 B | – | 30 layers | 32 heads | 4,096 tokens |
|---|
| DeepSeek DeepSeek-Coder-V2-Lite-Instruct | 15.7 B | – | 27 layers | 16 heads | 163,840 tokens |
|---|
| DeepSeek deepseek-moe-16b-base | 16.4 B | – | 28 layers | 16 heads | 4,096 tokens |
|---|
| DeepSeek deepseek-moe-16b-chat | 16.4 B | – | 28 layers | 16 heads | 4,096 tokens |
|---|
| DeepSeek DeepSeek-R1 | 671 B | 37 B | 61 layers | 128 heads | 163,840 tokens |
|---|
| DeepSeek DeepSeek-R1-0528 | 685 B | – | 61 layers | 128 heads | 163,840 tokens |
|---|
| DeepSeek DeepSeek-R1-0528-Qwen3-8B | 8.19 B | – | 36 layers | 8 heads | 131,072 tokens |
|---|
| DeepSeek DeepSeek-R1-Distill-Llama-8B | 8.03 B | – | 32 layers | 8 heads | 131,072 tokens |
|---|
| DeepSeek DeepSeek-R1-Distill-Llama-70B | 70.6 B | – | 80 layers | 8 heads | 131,072 tokens |
|---|
| DeepSeek DeepSeek-R1-Distill-Qwen-1.5B | 1.78 B | – | 28 layers | 2 heads | 131,072 tokens |
|---|
| DeepSeek DeepSeek-R1-Distill-Qwen-7B | 7.62 B | – | 28 layers | 4 heads | 131,072 tokens |
|---|
| DeepSeek DeepSeek-R1-Distill-Qwen-14B | 14.8 B | – | 48 layers | 8 heads | 131,072 tokens |
|---|
| DeepSeek DeepSeek-R1-Distill-Qwen-32B | 32.8 B | – | 64 layers | 8 heads | 131,072 tokens |
|---|
| DeepSeek DeepSeek-V2-Chat | 236 B | – | 60 layers | 128 heads | 163,840 tokens |
|---|
| DeepSeek DeepSeek-V2-Lite | 15.7 B | – | 27 layers | 16 heads | 163,840 tokens |
|---|
| DeepSeek DeepSeek-V2-Lite-Chat | 15.7 B | – | 27 layers | 16 heads | 163,840 tokens |
|---|
| DeepSeek DeepSeek-V3 | 671 B | 37 B | 61 layers | 128 heads | 163,840 tokens |
|---|
| DeepSeek DeepSeek-V3-0324 | 685 B | – | 61 layers | 128 heads | 163,840 tokens |
|---|
| DeepSeek DeepSeek-V3.2 | 685 B | – | 61 layers | 128 heads | 163,840 tokens |
|---|
| DeepSeek DeepSeek-V3.2-Exp | 685 B | – | 61 layers | 128 heads | 163,840 tokens |