← All Models
|
Qwen - Alibaba Cloud's multilingual LLM series:
Qwen1.5-0.5B
Qwen1.5-1.8B
Qwen1.5-4B
Qwen1.5-7B
Qwen1.5-14B
Qwen1.5-32B
Qwen1.5-72B
Qwen1.5-110B
Qwen1.5-MoE-A2.7B
Qwen2-0.5B
Qwen2-1.5B
Qwen2-7B
Qwen2-57B-A14B
Qwen2-72B
Qwen2.5-0.5B
Qwen2.5-1.5B
Qwen2.5-3B
Qwen2.5-7B
Qwen2.5-14B
Qwen2.5-32B
Qwen2.5-72B
Qwen3-0.6B
Qwen3-1.7B
Qwen3-4B
Qwen3-8B
Qwen3-14B
Qwen3-32B
Qwen3-30B-A3B
Qwen3-235B-A22B
Qwen/Qwen1.5-MoE-A2.7B
📊 Model Parameters
Total Parameters
14,315,784,192
Context Length
8,192
Hidden Size
2048
Layers
24
Attention Heads
16
KV Heads
16
💾 Memory Requirements
FP32 (Full)
53.33 GB
FP16 (Half)
26.67 GB
INT8 (Quantized)
13.33 GB
INT4 (Quantized)
6.67 GB
🔑 KV Cache (Inference)
Per Token (FP16)
196.61 KB
Max Context FP32
3.00 GB
Max Context FP16
1.50 GB
Max Context INT8
768.0 MB
⚙️ Model Configuration
Core Architecture
Vocabulary Size
151,936
Hidden Size
2,048
FFN Intermediate Size
5,632
Number of Layers
24
Attention Heads
16
KV Heads
16
Context & Position
Max Context Length
8,192
Uses Sliding Window
No
Sliding Window Size
Not set
Window Attention Layers
21
RoPE Base Frequency
1000000.0
RoPE Scaling
Not set
Attention Configuration
Attention Dropout
0%
Attention Bias
Yes
Tied Embeddings
No
Mixture of Experts
MoE Layer Frequency
1
Expert FFN Size
1,408
Shared Expert FFN Size
5,632
Experts per Token
4
Number of Experts
60
Normalize TopK Probabilities
No
Activation & Normalization
Activation Function
silu
RMSNorm Epsilon
1e-06
Special Tokens
BOS Token ID
151,643
Pad Token ID
Not set
EOS Token ID
151643
Data Type
Model Dtype
bfloat16
Layer Types:
Attention
MLP/FFN
Normalization
Embedding
Attention
MLP
Norm
Embedding
Clear
Expand All
Collapse All