deepseek 的网站图标

DeepSeek: DeepSeek V4.1 Flash 1017 (batch)

deepseek/deepseek-v4.1-flash-1017:batch

模型权重

DeepSeek V4.1 Flash 1017 is the October 17, 2026 snapshot of DeepSeek V4.1 Flash, served by DeepSeek under the deepseek-flash API model name. V4.1 Flash is a sparse mixture-of-experts model from DeepSeek, and the first built on the company's Causal Encoder-Decoder (CED) architecture. It activates 8B parameters on input and 16B on output from a 552B-parameter backbone, an asymmetric split that keeps per-token compute low relative to the model's total size. Image understanding is native to the architecture, with visual and text embeddings trained jointly from the start of pre-training rather than added afterward as in the earlier experimental V4 Flash Vision Exp.

It is suited for coding, terminal, and computer-use agents, along with long-horizon tasks that must run to completion across many steps and long-context analysis. Compressed KV caching cuts cache memory to roughly a quarter of the previous Flash generation, significantly reducing costs on agentic workloads. DeepSeek positions it as the cost-efficient tier of the V4.1 family and reports that it exceeds V4 Pro on performance, speed, and task completion time.

模态
输入/输出价格
US$0.036 / US$0.15每 1M
上下文
1.0M
发布时间
2026年10月17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-BenchTenstorrent92.8%--Baseten90.6%--PrimaLabs90.1%--AtlasCloud90.4%77.8%Morph91.1%76.7%Baseten90.0%76.5%Morph90.4%74.2%
+38 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: DeepSeek: DeepSeek V4.1 Flash 1017 (batch) (Artificial Analysis、Design Arena 和 Vals AI)
来源基准测试得分
Artificial AnalysisDeepSeek V4.1 Flash 1017 (High) Intelligence Index28.6
Artificial AnalysisDeepSeek V4.1 Flash 1017 (High) Terminal-Bench 4.028.8%
Artificial AnalysisDeepSeek V4.1 Flash 1017 (High) AA-Omniscience Accuracy38.9%
Artificial AnalysisDeepSeek V4.1 Flash 1017 (High) AA-Omniscience Non-Hallucination Rate38.8%
Artificial AnalysisDeepSeek V4.1 Flash (Max) HLE39.2%
Artificial AnalysisDeepSeek V4.1 Flash (Max) AA-LCR84.0%
Artificial AnalysisDeepSeek V4.1 Flash (Max) GDPval-AA55.0%
Artificial AnalysisDeepSeek V4.1 Flash (Max) CritPt14.3%
Artificial AnalysisDeepSeek V4.1 Flash (Max) SciCode51.9%
Artificial AnalysisDeepSeek V4.1 Flash (Max) Terminal-Bench 4.026.8%
Artificial AnalysisDeepSeek V4.1 Flash 1017 (Non-reasoning) Intelligence Index18.5
Artificial AnalysisDeepSeek V4.1 Flash 1017 (Non-reasoning) HLE11.5%
Artificial AnalysisDeepSeek V4.1 Flash 1017 (Non-reasoning) AA-LCR57.3%
Artificial AnalysisDeepSeek V4.1 Flash 1017 (Non-reasoning) CritPt0.3%
Artificial AnalysisDeepSeek V4.1 Flash 1017 (Non-reasoning) SciCode38.5%
Artificial AnalysisDeepSeek V4.1 Flash 1017 (Non-reasoning) Terminal-Bench 4.09.1%
Artificial AnalysisDeepSeek V4.1 Flash 1017 (Non-reasoning) AA-Omniscience Accuracy21.9%
Artificial AnalysisDeepSeek V4.1 Flash 1017 (Non-reasoning) AA-Omniscience Non-Hallucination Rate74.7%
Design ArenaDeepSeek-V4.1-Flash (1017) Agents Arena Spreadsheet Elo1265
Design ArenaDeepSeek-V4.1-Flash (1017) Models Arena 3D Elo1344
Design ArenaDeepSeek-V4.1-Flash (1017) Models Arena Asciiart Elo1219
Design ArenaDeepSeek-V4.1-Flash (1017) Models Arena Code Categories Elo1286
Design ArenaDeepSeek-V4.1-Flash (1017) Models Arena Data Visualization Elo1276
Design ArenaDeepSeek-V4.1-Flash (1017) Models Arena Game Development Elo1253
Design ArenaDeepSeek-V4.1-Flash (1017) Models Arena SVG Elo1241
Design ArenaDeepSeek-V4.1-Flash (1017) Models Arena UI Component Elo1280
Design ArenaDeepSeek-V4.1-Flash (1017) Models Arena Website Elo1293
Vals AIDeepSeek V4.1 Flash 1017 Access to Justice Benchmark73.1%
Vals AIDeepSeek V4.1 Flash 1017 Code Migration v1.167.5%
Vals AIDeepSeek V4.1 Flash 1017 CyberOffense54.8%
Vals AIDeepSeek V4.1 Flash 1017 EMB v1.169.2%
Vals AIDeepSeek V4.1 Flash 1017 Finance Agent v2.162.9%
Vals AIDeepSeek V4.1 Flash 1017 Vals Finance Index60.9%
Vals AIDeepSeek V4.1 Flash 1017 HLAB-Vals-1.114.2%
Vals AIDeepSeek V4.1 Flash 1017 Vals Legal Index71.9%
Vals AIDeepSeek V4.1 Flash 1017 Legal Research Bench45.2%
Vals AIDeepSeek V4.1 Flash 1017 Tax Agent Bench44.4%
Vals AIDeepSeek V4.1 Flash 1017 Terminal-Bench 4.031.8%
Vals AIDeepSeek V4.1 Flash 1017 Vals Index61.3%
Vals AIDeepSeek V4.1 Flash 1017 Vibe Code Bench 1-100 v1.126.2%
Vals AIDeepSeek V4.1 Flash 1017 Vibe Code Bench v1.187.7%

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。

DeepSeek: DeepSeek Flash 版本

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 DeepSeek: DeepSeek V4.1 Flash 1017 (batch) 构建应用的指南、实用教程、博客文章和工具。

常见问题