DeepSeek · 2023-11 · Dense
DeepSeek LLM 7B
DeepSeek LLM 7B is a multi-head attention (MHA) transformer released by DeepSeek in 2023-11, with 30 layers, hidden size 4096 and a context window of 4,096 tokens.
Layer stack
Attention ×30 · Layers 30
Key facts
| Family | DeepSeek |
|---|---|
| Released | 2023-11 |
| Params | 5.9 B |
| Context | 4,096 tokens |
| Attention | MHA (32:32) |
| Layers | 30 |
| Hidden | 4,096 |
| Heads | 32 |
| Vocab | 102,400 |
| Position | RoPE |
| Norm | RMSNorm |
| Activation | silu |
| Dtype | bf16 |
| Architecture class | LlamaForCausalLM |
Architecture overview
Attention
multi-head attention (MHA) — 32 q-heads / 32 kv-heads. RoPE θ=10,000.
Feed-forward / MoE
It is a dense model: all 5.9B parameters participate in computing every token.
Field-level comparison
Compared against the previous model of the same lab; where none exists, against the structurally closest model. The ratio column is this model divided by the comparison model.
Full-field comparison vs predecessor DeepSeek-Coder 6.7B
| Model | DeepSeek LLM 7B | DeepSeek-Coder 6.7B | Ratio |
|---|---|---|---|
| model_type | llama | llama | ＝ |
| architectures | LlamaForCausalLM | LlamaForCausalLM | ＝ |
| hidden_size | 4096 | 4096 | ≈1 |
| num_hidden_layers | 30 | 32 | ×0.938 |
| num_attention_heads | 32 | 32 | ≈1 |
| num_key_value_heads | 32 | 32 | ≈1 |
| intermediate_size | 11008 | 11008 | ≈1 |
| hidden_act | silu | silu | ＝ |
| rope_theta | 10000 | 100000 | ×0.100 |
| max_position_embeddings | 4096 | 16384 | ×0.250 |
| rms_norm_eps | 0.000001 | 0.000001 | ≈1 |
| vocab_size | 102400 | 32256 | ×3.17 |
| torch_dtype | bfloat16 | bfloat16 | ＝ |
| bos_token_id | 1 | 32013 | ×0.000 |
| eos_token_id | 2 | 32014 | ×0.000 |
| initializer_range | 0.02 | 0.02 | ≈1 |
| pretraining_tp | 1 | 1 | ≈1 |
| tie_word_embeddings | false | false | ＝ |
| transformers_version | 4.33.1 | 4.34.1 | ≠ |
| use_cache | true | true | ＝ |
Most similar architectures
- DeepSeek-Coder 6.7B (DeepSeek, 2023-11) 0.999
- Qwen1.5 7B (Qwen, 2024-02) 0.998
- LLaMA 1 7B (Meta, 2023-02) 0.882
- DeepSeekMoE 16B (DeepSeek, 2024-01) 0.874
- Qwen3 8B (Qwen, 2025-04) 0.782
- Hunyuan 7B (Hunyuan, 2025-08) 0.782
Similarity values range from 0 (no shared categorical features) to 1 (identical profiles).
Raw config fields
20 fields
| architectures | LlamaForCausalLM |
|---|---|
| bos_token_id | 1 |
| eos_token_id | 2 |
| hidden_act | silu |
| hidden_size | 4096 |
| initializer_range | 0.02 |
| intermediate_size | 11008 |
| max_position_embeddings | 4096 |
| model_type | llama |
| num_attention_heads | 32 |
| num_hidden_layers | 30 |
| num_key_value_heads | 32 |
| pretraining_tp | 1 |
| rms_norm_eps | 0.000001 |
| rope_theta | 10000 |
| tie_word_embeddings | false |
| torch_dtype | bfloat16 |
| transformers_version | 4.33.1 |
| use_cache | true |
| vocab_size | 102400 |