LLM
Qwen3.8-Flash-Next
Qwen3.8-Flash-NextQwen's 125B sparse MoE vision-language model (6B active) for coding, reasoning, and agents. Text/image in; text out; 262K native context.
Model details
View repositoryQwen3.8-Flash-Next is Qwen’s open-weight vision-language model for coding, reasoning, and agentic workflows. It uses a 125B sparse Mixture-of-Experts (MoE) language model with 6B active parameters, a vision encoder, a 51B n-gram embedding table, and one multi-token prediction layer. Its hybrid design combines Gated DeltaNet with Qwen Sparse Attention.
The model accepts text and image inputs and generates text. It supports thinking and tool use, with a native context of 262,144 tokens.
