88Z AI, 27 August — Alibaba Tongyi has released Qwen3.8-Flash, a multimodal MoE model with 125B total parameters and 6B activated per token, supporting ultra-long contexts. With four major technical upgrades, training costs are significantly reduced, coding and office capabilities are enhanced, and competitive API pricing is offered.
At a glance
- The model has 125B total parameters with 6B activated per token, natively supporting 260k context extendable to 1M.
- Four technical upgrades reduce training costs to 1/9 of the previous generation, with API pricing at 1 yuan input and 3 yuan output per million tokens.
The update
Alibaba Tongyi has officially launched Qwen3.8-Flash, a multimodal Mixture of Experts model. It features a total of 125 billion parameters, with only 6 billion activated per token during inference, enabling efficient computation. The model natively supports a context window of 260,000 tokens and can be extended to one million tokens. Architecturally, it incorporates four key upgrades: GDN and QSA hybrid attention mechanisms, gated residual connections, N-gram embeddings, and the Muon optimizer. These improvements reduce training costs to one-ninth of the previous generation while significantly enhancing code generation and office task processing capabilities. The model is now available via API, priced at 1 yuan per million tokens for input and 3 yuan per million tokens for output.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- 千问大模型Source published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.