qwen 的网站图标

Qwen: Qwen3.8 27B (batch)

qwen/qwen3.8-27b:batch

模型权重

Qwen3.8 27B is an open-weight dense vision-language model from Qwen. It is suited for coding, professional workflows, research, multimodal interaction, and long-running agent tasks, with flexible thinking that can be enabled or disabled.

模态
输入/输出价格
US$0.16 / US$2每 1M
上下文
262K
发布时间
2026年8月14日

提供商

不同公司托管同一模型。OpenRouter 根据您选择的路由模式将请求路由到其中一家公司:Balanced(价格 + 速度)、Nitro(速度最快)或 Floor(价格最低)。

定价

客户为此模型实际支付的平均价格,与提供商公布的价格并列展示。由于缓存和折扣,实际支付的价格往往远低于标价。

加权平均值

输入
US$0/百万 token
输出
US$0/百万 token
价格历史
当前选择无实际价格数据

性能

吞吐量是模型生成内容的速度(每秒 token 数,越高越好)。延迟是完整的往返时间(越低越好)。TTFT 是首个 token 的响应时间,即您看到内容出现前需要等待的时间(越低越好)。

AutoExacto 基准测试
GPQA DiamondTAU-BenchStreamLake84.9%--GMICloud82.1%--DekaLLM83.3%77.9%Parasail83.9%77.3%PrimeIntellect83.2%78.0%
+21 家其他提供商

正常运行时间

正常运行时间是指过去 3 天中至少有一家提供商响应请求的时间百分比。可用性是指成功提供推理服务的时间百分比。OpenRouter 会持续监控,并在某家提供商返回错误时使用次优提供商。

正常运行时间数据尚不足,暂时无法显示。

当上游提供商发生错误时,如果您的请求筛选条件允许,我们可以通过路由到另一家健康的提供商来恢复服务。您可以通过以下接口以编程方式访问各提供商的正常运行时间数据: 端点 API。了解更多有关我们的 负载均衡和提供商路由选项。

基准测试

标准化评测得分,以及此模型针对共用基准测试提示词的实际输出。百分比越高越好,每项排名显示此模型在该来源所评测的模型中的位置。

基准测试得分摘要: Qwen: Qwen3.8 27B (batch) (Artificial Analysis 和 Vals AI)
来源基准测试得分
Artificial AnalysisQwen3.8 27B (Xhigh) Intelligence Index20.1
Artificial AnalysisQwen3.8 27B (Xhigh) Coding Index68.1
Artificial AnalysisQwen3.8 27B (Xhigh) Agentic Index45.8
Artificial AnalysisQwen3.8 27B (Xhigh) GPQA Diamond90.5%
Artificial AnalysisQwen3.8 27B (Xhigh) HLE33.9%
Artificial AnalysisQwen3.8 27B (Xhigh) AA-LCR82.0%
Artificial AnalysisQwen3.8 27B (Xhigh) τ-Bench Banking48.0%
Artificial AnalysisQwen3.8 27B (Xhigh) GDPval-AA46.3%
Artificial AnalysisQwen3.8 27B (Xhigh) CritPt5.4%
Artificial AnalysisQwen3.8 27B (Xhigh) SciCode46.6%
Artificial AnalysisQwen3.8 27B (Xhigh) Terminal-Bench 4.05.6%
Artificial AnalysisQwen3.8 27B (Xhigh) Terminal-Bench 2.179.8%
Artificial AnalysisQwen3.8 27B (Xhigh) AA-Omniscience Accuracy17.3%
Artificial AnalysisQwen3.8 27B (Xhigh) AA-Omniscience Non-Hallucination Rate71.4%
Artificial AnalysisQwen3.8 27B (Low) Intelligence Index15.7
Artificial AnalysisQwen3.8 27B (Low) Coding Index58.2
Artificial AnalysisQwen3.8 27B (Low) Agentic Index38.9
Artificial AnalysisQwen3.8 27B (Low) GPQA Diamond84.5%
Artificial AnalysisQwen3.8 27B (Low) HLE14.0%
Artificial AnalysisQwen3.8 27B (Low) AA-LCR77.3%
Artificial AnalysisQwen3.8 27B (Low) τ-Bench Banking32.2%
Artificial AnalysisQwen3.8 27B (Low) GDPval-AA43.3%
Artificial AnalysisQwen3.8 27B (Low) CritPt0.0%
Artificial AnalysisQwen3.8 27B (Low) SciCode40.0%
Artificial AnalysisQwen3.8 27B (Low) Terminal-Bench 4.02.5%
Artificial AnalysisQwen3.8 27B (Low) Terminal-Bench 2.167.4%
Artificial AnalysisQwen3.8 27B (Low) AA-Omniscience Accuracy18.3%
Artificial AnalysisQwen3.8 27B (Low) AA-Omniscience Non-Hallucination Rate51.3%
Artificial AnalysisQwen3.8 27B (Non-reasoning) Intelligence Index10.0
Artificial AnalysisQwen3.8 27B (Non-reasoning) Coding Index44.6
Artificial AnalysisQwen3.8 27B (Non-reasoning) Agentic Index22.4
Artificial AnalysisQwen3.8 27B (Non-reasoning) GPQA Diamond81.8%
Artificial AnalysisQwen3.8 27B (Non-reasoning) HLE12.1%
Artificial AnalysisQwen3.8 27B (Non-reasoning) AA-LCR69.3%
Artificial AnalysisQwen3.8 27B (Non-reasoning) τ-Bench Banking20.0%
Artificial AnalysisQwen3.8 27B (Non-reasoning) GDPval-AA29.1%
Artificial AnalysisQwen3.8 27B (Non-reasoning) CritPt0.3%
Artificial AnalysisQwen3.8 27B (Non-reasoning) SciCode36.2%
Artificial AnalysisQwen3.8 27B (Non-reasoning) Terminal-Bench 4.00.0%
Artificial AnalysisQwen3.8 27B (Non-reasoning) Terminal-Bench 2.149.1%
Artificial AnalysisQwen3.8 27B (Non-reasoning) AA-Omniscience Accuracy9.1%
Artificial AnalysisQwen3.8 27B (Non-reasoning) AA-Omniscience Non-Hallucination Rate82.0%
Artificial AnalysisQwen3.8 27B (Medium) Intelligence Index17.5
Artificial AnalysisQwen3.8 27B (Medium) Coding Index56.1
Artificial AnalysisQwen3.8 27B (Medium) Agentic Index44.4
Artificial AnalysisQwen3.8 27B (Medium) GPQA Diamond84.5%
Artificial AnalysisQwen3.8 27B (Medium) HLE14.1%
Artificial AnalysisQwen3.8 27B (Medium) AA-LCR79.7%
Artificial AnalysisQwen3.8 27B (Medium) τ-Bench Banking47.4%
Artificial AnalysisQwen3.8 27B (Medium) GDPval-AA44.5%
Artificial AnalysisQwen3.8 27B (Medium) CritPt0.0%
Artificial AnalysisQwen3.8 27B (Medium) SciCode39.0%
Artificial AnalysisQwen3.8 27B (Medium) Terminal-Bench 4.05.1%
Artificial AnalysisQwen3.8 27B (Medium) Terminal-Bench 2.165.2%
Artificial AnalysisQwen3.8 27B (Medium) AA-Omniscience Accuracy17.9%
Artificial AnalysisQwen3.8 27B (Medium) AA-Omniscience Non-Hallucination Rate34.6%
Vals AIQwen 3.8 27B Code Migration14.2%
Vals AIQwen 3.8 27B CyberBench v1.167.8%
Vals AIQwen 3.8 27B EMB59.7%
Vals AIQwen 3.8 27B Finance Agent (v2)48.6%
Vals AIQwen 3.8 27B GPQA Diamond88.9%
Vals AIQwen 3.8 27B Harvey's Legal Agent Benchmark v111.3%
Vals AIQwen 3.8 27B IOI39.1%
Vals AIQwen 3.8 27B LiveCodeBench84.0%
Vals AIQwen 3.8 27B LegalBench82.4%
Vals AIQwen 3.8 27B Legal Research Bench36.1%
Vals AIQwen 3.8 27B MedCode28.7%
Vals AIQwen 3.8 27B MedScribe83.8%
Vals AIQwen 3.8 27B MMLU Pro84.3%
Vals AIQwen 3.8 27B MMMU Pro83.9%
Vals AIQwen 3.8 27B MortgageTax64.9%
Vals AIQwen 3.8 27B ProgramBench0.0%
Vals AIQwen 3.8 27B ProofBench v1.116.0%
Vals AIQwen 3.8 27B SAGE v152.4%
Vals AIQwen 3.8 27B SkillsBench38.1%
Vals AIQwen 3.8 27B SWE-bench86.0%
Vals AIQwen 3.8 27B Tax Agent Bench30.9%
Vals AIQwen 3.8 27B TaxEval v270.8%
Vals AIQwen 3.8 27B Terminal-Bench 2.158.4%
Vals AIQwen 3.8 27B Terminal-Bench-Science0.0%
Vals AIQwen 3.8 27B Vals Index48.5%
Vals AIQwen 3.8 27B Vibe Code Bench v1.164.8%

应用

向此模型发送最多流量的公开应用。这能很好地反映实际生产工作负载的情况,也能提示此模型最适合哪些用例。

使用此模型的应用尚无足够数据。

活动

此模型的 token 用量和请求流量随时间的变化。

API

可直接用于调用此模型的代码。OpenRouter 的 API 与 OpenAI 兼容——大多数 SDK 只需替换基础 URL 即可使用。切换模型时,唯一需要更改的是下方的模型标识符。

目前无法通过 EU 区域内路由使用。 区域内路由文档

文档

在 OpenRouter 上使用 Qwen: Qwen3.8 27B (batch) 构建应用的指南、实用教程、博客文章和工具。

常见问题