Qwen 3.8 Flash Next
Active Parameters
124B
Context Length
262K
Modality
Text
Architecture
Mixture of Experts (MoE)
License
other
Release Date
24 Aug 2026
Knowledge Cutoff
-
System Requirements
VRAM requirements for different quantization methods and context sizes
1,024 tokens
Consumer
13x RTX 4090
24GB VRAM
Datacenter
4x NVIDIA A100
80GB VRAM
Apple Silicon
3x Apple M3 Max
128GB VRAM
262,144 tokens
Consumer
15x RTX 4090
24GB VRAM
Datacenter
4x NVIDIA A100
80GB VRAM
Apple Silicon
3x Apple M3 Max
128GB VRAM
Architecture Diagram
Evaluation Benchmarks
Rank
#42
| Benchmark | Score | Rank |
|---|---|---|
Agentic Coding | 0.62 | ⭐ 8 |
Web Development | 1622 | ⭐ 8 |
Reasoning | 0.87 | 20 |
General | 0.76 | 22 |
LiveBench Average | 0.76 | 26 |
Data Analysis | 0.74 | 32 |
Agent Arena | 0.03 | 40 |
Coding | 0.73 | 48 |
Mathematics | 0.86 | 53 |
Rankings
Overall Rank
#42
Coding Rank
#52
About Qwen 3.8 Flash Next
Qwen 3.8 Flash Next is an experimental next-generation language model optimized for ultra-fast inference and efficient reasoning. It incorporates Qwen's latest architectural improvements to deliver high throughput across complex tasks.
Technical Specifications
Attention
Attention Structure
DeepSeek Sparse Attention
Attention Heads
24
Key-Value Heads
2
Attention Head Dimension
256
Position Embedding
ROPE
RoPE Theta
10,000,000
Sliding Window Attention
No
Sliding Window Size
-
Sliding Window Ratio
-
Linear Attention
Yes
Linear Attention Ratio
75.0%
Normalization
RMS Normalization
Activation Function
SwigLU
Dimensions
Hidden Dimension Size
2,560
Number of Layers
48
FFN Intermediate Size (Dense)
-
Multi-Token Prediction Heads
1
Tokenizer
Vocabulary Size
248,320
Mixture of Experts
Total Expert Parameters
5.5B
Number of Experts
512
Active Experts
10
Shared Experts
1
FFN Intermediate Size (per Expert)
640
Dense Layers Before MoE
-
Resources
About Qwen 3.8
Alibaba's Qwen 3.8 generation represents the frontier hybrid Mixture-of-Experts architecture designed for coding, professional work, research, and long-horizon agentic tasks. It features a 2.4-trillion parameter architecture (95B active per token) combining Gated DeltaNet linear attention with standard Gated Attention, available both as open weights and as a hosted flagship service.