Qwen3.8-Flash-Next
Qwen3.8-Flash-Next introduces an efficiency-oriented architecture for multimodal reasoning and agents. It combines Gated DeltaNet with Qwen Sparse Attention, plus 51B n-gram embedding parameters held outside the accelerator and 4B multi-token-prediction parameters. The 125B backbone activates 6B parameters per token. Public weights use the Qwen Community License 1.0 and support a native 256K context with optional million-token extension. The hosted Qwen3.8-Flash service adds production features and has a separate page.
2026-08-26
125B backbone + 51B n-gram embeddings + 4B MTP; 6B activated backbone
Mixture of Experts (MoE) with Gated DeltaNet, Qwen Sparse Attention, n-gram memory, and vision encoder
Qwen Community License 1.0
Specifications
- Parameters
- 125B backbone + 51B n-gram embeddings + 4B MTP; 6B activated backbone
- Architecture
- Mixture of Experts (MoE) with Gated DeltaNet, Qwen Sparse Attention, n-gram memory, and vision encoder
- License
- Qwen Community License 1.0
- Context Window
- 262,144 tokens
- Type
- multimodal
- Modalities
- textimagevideo
Benchmark Scores
Advanced Specifications
- Model Family
- Qwen
- API Access
- Not Available
- Chat Interface
- Not Available
Capabilities & Limitations
- Capabilities
- reasoningcode generationtool usevisual reasoningefficient long-context inference
- Known Limitations
- 125B backbone count excludes 51B n-gram embeddings and 4B MTP parameters.Native context is 262,144 tokens; hosted Flash uses different production limits.Custom license conditions apply.
- Notable Use Cases
- coding assistantsresearch agentsdocument analysismultimodal agents
- Function Calling Support
- Yes
- Tool Use Support
- Yes