Overview
Qwen3.8-Flash-Next is an open-weight multimodal mixture-of-experts model and an early preview of the architecture intended for Qwen4. Its main model has 125 billion parameters, with 6 billion activated per token, plus 51 billion N-gram embedding parameters. Its architecture combines Gated DeltaNet with Qwen Sparse Attention and uses gated residual connections. (alibabacloud.com)
Capabilities and uses
The model supports language and visual understanding and is intended for tasks including coding, office work and long-context processing. It supports 262,144 tokens of context natively, extendable to 1,000,000 tokens with YaRN. Alibaba also offers a distinct production version, Qwen3.8-Flash, with a 1-million-token context by default and built-in tools. (alibabacloud.com)
Limitations
The open-weight model is an architectural preview; extending its context beyond the native length requires an appropriate configuration. Self-hosting requires suitable compute and memory resources. The model weights are distributed under the Qwen Community 1.0 license; users should review its terms. Outputs should be checked for accuracy before consequential use. (huggingface.co)
Pricing
No separate official per-token API price for the open-weight Qwen3.8-Flash-Next was verified. The prices below are for the related production service Qwen3.8-Flash, not a claim that the two offerings are identical. They are original invocation prices, excluding promotions. (huggingface.co)
- China domestic site, Model Studio, Beijing: CNY 0.8 per million input tokens and CNY 2.7 per million output tokens. Cache-hit input costs CNY 0.1 per million tokens; explicit cache creation costs CNY 1.25 per million tokens. The published standard rate covers requests with more than 0 and up to 1 million input tokens, in thinking and non-thinking modes. Source: Alibaba Cloud China pricing, checked November 16, 2026. (help.aliyun.com)
- International site, Model Studio, Singapore international deployment: USD 0.15 per million input tokens and USD 0.47 per million output tokens. Implicit cache-hit input and explicit cache reads each cost USD 0.016 per million tokens; explicit cache creation costs USD 0.2 per million tokens. The standard rate covers requests with more than 0 and up to 1 million input tokens. Source: Alibaba Cloud international pricing, checked November 16, 2026. (alibabacloud.com)
These are different regional service offerings, not currency conversions of one package. Other deployment scopes and regions have different prices. (help.aliyun.com)
Release and official sources
- Release date: August 26, 2026, according to the official repository. (github.com)
- Official GitHub repository
- Official model weights and model card
- Alibaba Cloud announcement