PulseAugur
中
实时 08:47:10
中文(ZH) 全新架构，极致性价比！阿里千问Qwen3.8-Flash发布并开源

阿里发布Qwen3.8-Flash-Next，成本大降、性能飙升 · 追踪6个来源

作者 PulseAugur 编辑部 · [9 个来源] ·

阿里已发布并开源Qwen3.8-Flash-Next，这是一款多模态的混合专家模型，预览了即将推出的Qwen4架构。该新模型拥有1250亿总参数，但每个token仅激活60亿参数，显著降低了训练和推理成本。据报道，Qwen3.8-Flash-Next在包括编码和办公任务在内的多项基准测试中，性能超越了Claude Opus 4.6和DeepSeek-V4-Flash等竞争对手，同时API定价也大幅降低。 AI

影响 为大语言模型的成本效益设定了新标杆，可能在定价和性能方面给OpenAI和Anthropic等竞争对手带来压力。

排序理由 前沿实验室模型发布，附带系统卡。

在 雷峰网 (Leiphone) 阅读 →

AI 生成摘要 · Google Gemini · 来自 9 个来源。 我们如何撰写摘要 →

阿里发布Qwen3.8-Flash-Next，成本大降、性能飙升 · 追踪6个来源

本文如何被排名

Signal score
1 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Frontier Release
前沿实验室模型发布，附带系统卡。
Source corroboration
9 independent sources
Strong cross-source corroboration — multiple independent publishers covered this within the clustering window.
Topics
model release, product
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
5 days old
Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.
Coverage growth since scoring
+3 source(s) since last score
New sources have picked up this story since our last re-score. Score will update on the next scoring pass.

完整方法见我们的编辑标准。

报道来源 [9]

  1. X — SemiAnalysis TIER_1 English(EN) · SemiAnalysis_ ·

    祝贺 @Alibaba_Qwen 发布 Qwen3.8-Flash-Next，采用了与其即将发布的 Qwen4 模型相同的架构创新！此类创新包括：

    Congrats to @Alibaba_Qwen on the release of Qwen3.8-Flash-Next, using the same architecture innovations as their upcoming Qwen4 model! Such innovations include: 🟠 51-billion-param N-gram Embedding to look up a table with very little extra computation, which means the embedding h…

  2. 雷峰网 (Leiphone) TIER_1 中文(ZH) ·

    全新架构，极致性价比！阿里云 Qwen3.8-Flash 发布并开源

    <p>8月26日晚，阿里发布并同步开源千问最新模型 Qwen3.8-Flash。该模型采用全新下一代架构，千亿总参数仅激活60亿（6B）即可获得超越Claude Opus4.6的前沿性能，创下模型效率的全球新基准。得益于架构和训练的全面革新，Qwen3.8-Flash训练成本较Qwen3.7-Plus骤降近90%，推理成本同样大幅下降，每百万Tokens输入仅1元，输出3元，价格最低至DeepSeek-V4-Flash的1/3，将全球模型性价比“斩杀”至极致新水平。Qwen3.8-Flash今晚将首发上线“千问办公”，开发者和企业也可通过千问AI平台获取…

  3. The Decoder TIER_1 English(EN) · Matthias Bastian ·

    阿里巴巴发布 Qwen3.8-Flash-Next，目标是“极致的成本效益”

    <p><img alt="" class="attachment-full size-full wp-post-image" height="1152" src="https://the-decoder.com/wp-content/uploads/2026/07/qwen_logo-1.png" style="height: auto; margin-bottom: 10px;" width="2048" /></p> <p> Alibaba's Qwen team is previewing the Qwen4 architecture with Q…

  4. Pandaily TIER_1 English(EN) · [email protected] (Pandaily) ·

    阿里巴巴开源 Qwen3.8-Flash，拥有 60 亿活跃参数，训练成本仅为九分之一

    Alibaba's Qwen team released Qwen3.8-Flash, a multimodal MoE preview of the Qwen4 architecture with only 6B activated parameters and a training cost one-ninth that of Qwen3.7-Plus.

  5. Mastodon — sigmoid.social TIER_1 Italiano(IT) · [email protected] ·

    阿里巴巴推出 Qwen3.8-Flash：这款 27B 模型能像 Claude 一样编写代码。付费 AI 的未来又将如何？📌 文章链接：https://www.redhotcyber.com/

    Alibaba lancia Qwen3.8-Flash: il modello da 27B scrive codice come Claude. E il futuro delle IA a pagamento? 📌 Link all'articolo : https://www. redhotcyber.com/post/alibaba-l ancia-qwen3-8-flash-il-modello-da-27b-scrive-codice-come-claude-e-il-futuro-delle-ia-a-pagamento/ Massimi…

    链接 redhotcyber.com/…/alibaba-lancia-qwen3-8-… redhotcyber.com/…/alibaba-lancia-occhiali…
  6. Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] ·

    阿里巴巴开源了 Qwen3.8-Flash，一个拥有1250亿参数但每个 token 只激活60亿的混合专家多模态模型。训练成本

    Alibaba has open-sourced Qwen3.8-Flash, a multimodal mixture-of-experts model with 125 billion parameters but only 6 billion activated per token. Training cost is about one-ninth that of its predecessor, with API pricing at 1 yuan per million input tokens. https:// pandaily.com/a…

    链接 pandaily.com/alibaba-qwen3-8-flash-open-s…
  7. Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] ·

    阿里巴巴发布 Qwen3.8-Flash-Next，一款拥有仅 6B 激活参数的 125B 多模态混合专家模型。该发布预览了 Qwen4 架构和

    Alibaba has released Qwen3.8-Flash-Next, a 125B multimodal Mixture-of-Experts model with just 6B active parameters. The release previews Qwen4 architecture and reportedly trains at one-ninth the cost of Qwen3.7-Plus. Weights coming soon. https://www. marktechpost.com/2026/08/26/a…

  8. Mastodon — mastodon.social TIER_1 Italiano(IT) · [email protected] ·

    👀 阿里巴巴发布了 Qwen3.8-Flash-Next 并开放权重，510亿参数可载入系统内存而非GPU。这是预览版

    👀 Alibaba ha pubblicato Qwen3.8-Flash-Next con i pesi aperti, e 51 miliardi di quei parametri stanno nella RAM di sistema invece che sulla GPU. È un'anteprima di Qwen4, ed è già scaricabile su Hugging Face. 👇 https:// gomoot.com/qwen3-8-flash-next- e-lassaggio-di-qwen4-ed-e-gia-s…

    链接 gomoot.com/qwen3-8-flash-next-e-lassaggio…
  9. Mastodon — mastodon.social TIER_1 English(EN) · sipirtu ·

    阿里巴巴发布 Qwen3.8-Flash-Next，目标是极致的成本效益。该混合专家模型每个 token 只激活 1250 亿参数中的 6 个。

    Alibaba released Qwen3.8-Flash-Next, targeting ultimate cost efficiency. The mixture-of-experts model activates just 6 out of 125 billion parameters per token. Source: The Decoder AI https:// the-decoder.com/alibaba-releas es-qwen3-8-flash-next-targeting-ultimate-cost-efficiency/…

    链接 the-decoder.com/alibaba-releases-qwen3-8-…

相关实体

相关话题