AI News
官方更新 DeepSeek Blog 2026-10-17 正文 ok 2335 words

DeepSeek 开放自研搜索 API

英文标题：📋 Opening up the in-house search API​

DeepSeek 宣布开放其自研 DeepSeek Search 的 API，该搜索引擎每日稳定处理全球数亿次请求，索引数千亿网页，并支持多语言实时检索。公司称，该系统通过严格检测过滤 SEO 垃圾内容、GEO 操纵和低质量 AIGC，旨在为用户及 LLM 提供更具权威性和可信度的信息来源。此次更新还将 DeepSeek-V4.1-Flash 升级至 1017 版本，开发者继续使用 `deepseek-flash` 即可调用，API 调用方式和价格不变。完整更新日志同时梳理了 DeepSeek V4、V3.2、V3.1 及 R1 系列模型在多模态、Agent、Responses API、推理能力和定价方面的演进。
打开原文相关 4.8 · 机会 3
media3.5 KB media3.5 KB media62.7 KB media455.5 KB

中文

更新日志

日期：2026-10-17

开放自研搜索 API

DeepSeek 的网页端和 API 搜索能力由自研 DeepSeek 搜索引擎提供支持。该引擎每天稳定处理来自全球的数亿次搜索请求。DeepSeek Search 是一款 AI 原生的通用搜索引擎，索引数千亿网页，支持多语言搜索，并能够实时返回高相关性、权威且及时的结果。该引擎通过严格的检测机制过滤 SEO 垃圾内容、GEO 操纵和低质量 AIGC，确保为用户和 LLM 提供权威、可信的信息。使用详情请参阅 Search & Fetch 指南。

DeepSeek-V4.1-Flash-1017 更新

DeepSeek V4.1 Flash 已升级至 DeepSeek-V4.1-Flash-1017。官方网页端和 APP 端模型也已更新至新版本。API 调用方式和价格保持不变——只需使用 deepseek-flash 即可访问最新模型。

性能—成本帕累托改进

与此前的 DeepSeek V4.1 相比，新模型在各项基准测试中实现了显著提升：

  • HLE（不使用工具）：42.6（44.0*）

  • Terminal-Bench 4.0：42.3

  • DeepSWE v1.1：77.0

  • ProgramBench：29.1

  • NL2Repo-Bench：68.2

  • Automation-Bench v1.0.6：60.5

  • Agents' Last Exam：35.4

  • Apex Bench v1.1：72.1

  • Chartography（使用工具）：83.0

  • BabyVision（使用工具）：92.5

  • ZeroBench-main（使用工具）：51.0

  • 仅在 HLE 基准测试集的纯文本子集上进行测试。Terminal-Bench 4.0 使用 Claude Code 框架测试，DeepSWE 使用 Mini-SWE-V2 框架测试。ZeroBench-main 采用 Pass@5 评估。

在提升性能的同时，我们也降低了模型的任务运行时间和输出长度，使用户能够以更低成本、更快速度完成任务，并在多个测试框架中实现新的性能—成本帕累托最优。我们特别针对 DeepSeek Harness、腾讯 CodeBuddy、WorkBuddy 及其他平台进行了优化。为获得使用新模型的最佳体验，请升级至 DeepSeek Harness 0.2.1-rc.1 或更高版本。

日期：2026-09-10

DeepSeek-V4.1-Flash 发布

今日，我们正式发布 DeepSeek-V4.1-Flash 模型。它是我们新架构系列中尺寸最小的模型，具备原生多模态视觉理解能力。新架构旨在实现更高的能力上限、更快的推理速度、更高的吞吐量，并可扩展至更大规模的模型。

  • GPQA Diamond：90.9

  • HLE：36.8（39.1*）

  • Codeforces（Rating）：3471

  • MathArena Apex：65.6

  • Terminal-Bench 2.1：90.6

  • Terminal-Bench 3.0：30.0

  • Terminal-Bench 4.0：31.2

  • DeepSWE v1.1：74.2

  • ProgramBench：20.3

  • NL2Repo-Bench：65.4

  • CyberGym：88.1

  • SEC-Bench Pro：62.8

  • ExploitGym：15.3

  • HLE（使用工具）：63.9

  • Automation-Bench：54.8

  • Agents' Last Exam：31.8

  • Chartography（使用工具）：78.9

  • BabyVision（使用工具）：89.6

  • ZeroBench-main（使用工具）：49.0

  • 仅在 HLE 基准测试集的纯文本子集上进行测试。

API 变更

DeepSeek V4.1 Flash 现已在 DeepSeek API 上提供，并支持原生多模态能力。将模型名称改为 deepseek-flash，即可调用最新的 V4.1 Flash 模型。上一代模型 V4 Flash 和 V4 Flash Vision Exp 已退役；为保证兼容性，模型名称 deepseek-v4-flash 和 deepseek-v4-flash-vision-exp 将暂时路由至 V4.1 Flash。

根据用户需求，我们决定在 2026 年 9 月 14 日之后继续提供 DeepSeek V4 Pro 的 API 服务，计费方式维持不变。如有任何变更，我们将另行通知。感谢您的理解与支持！

API 定价调整

随着 DeepSeek-V4.1-Flash 的发布，API 价格也相应下调。详情请参阅 模型与定价。

更多详情请参阅 此文档。

日期：2026-08-21

DeepSeek-V4-Flash-Vision-Exp 发布

今日，新的多模态视觉理解模型 DeepSeek-V4-Flash-Vision-Exp 已在 DeepSeek API 平台上线。这是一款实验性模型，可通过设置 model='deepseek-v4-flash-vision-exp' 访问。

  • Terminal Bench 2.1：83.9

  • NL2Repo：57.7

  • DeepSWE：59.3

  • DSBench-Hard：63.6

  • AutomationBench（Public）：25.7

  • ApexBench（Pass@1）：36.5

  • Agents' Last Exam：27.3

  • Chartography：64.3

  • ZeroBench（Pass@5）：35.0

  • 对于公开基准测试集中的 Code Agent 文本任务，DeepSeek 系列模型均使用 DeepSeek Harness 最简模式作为测试框架，采用 max effort 级别、topp=0.95 和 temperature=1.0；在 ApexBench 和 Agents' Last Exam 评测中，文本模型 DeepSeek-V4-Flash 会忽略其中的多模态元素。

在纯文本能力方面（Agent、推理、世界知识等），DeepSeek-V4-Flash-Vision-Exp 与正式版 DeepSeek-V4-Flash 持平。

在需要视觉理解能力的 Agent 基准测试中，DeepSeek-V4-Flash-Vision-Exp 相比 DeepSeek-V4-Flash 实现显著跃升，其多模态 Agent 能力已接近 Opus-4.8。

使用详情请参阅 Vision 指南。

日期：2026-08-13

DeepSeek-V4-Pro 更新

DeepSeek-V4-Pro 的 GA 正式版本已在 APP、Web 和 API 上线。API 调用方式保持不变——只需将模型名称设置为 deepseek-v4-pro，即可使用最新版本。

Agent 能力显著增强

DeepSeek V4 Pro 的 GA 版本大幅增强了 Agent 能力，在生产环境中的性能提升尤其显著。

  • HLE（不使用工具 / 使用工具）：42.7/60.0
  • Terminal Bench 2.1：87.9
  • NL2Repo：61.5
  • Cybergym：83.3
  • DeepSWE：62.7
  • Toolathlon-Verified：74.1
  • Agents' Last Exam：25.7
  • AutomationBench（Public）：31.8
  • DSBench-FullStack：71.1
  • DSBench-Hard：67.2

原生支持 Responses API

DeepSeek API 现已原生支持 OpenAI Responses API 格式，并针对 Codex 进行了适配。用户可参阅官方文档，并通过一键配置脚本完成 Codex 配置。

更灵活的思考强度控制

V4-Pro 和 V4-Flash 的思考模式现支持三档思考强度：low / high / max。在实际使用中，用户可根据任务复杂度灵活选择：简单任务使用 low，日常 Agent 任务使用 high，更复杂的场景使用 max。设置说明请参阅官方 API 文档：思考模式。

API 定价调整

随着 DeepSeek V4 模型系列的正式发布，我们将更新并调整 API 定价。为更合理地分配资源，我们将采用高峰 / 非高峰定价，其中非高峰价格为高峰时段价格的一半，鼓励用户根据实际使用情况安排任务。新价格将于 2026 年 8 月 16 日 16:00（UTC 时间）生效。

更多详情请参阅 此文档。

日期：2026-07-31

DeepSeek-V4-Flash 更新

DeepSeek-V4-Flash API 正式版现已进入公开测试阶段。API 调用方式保持不变——只需将模型名称设置为 deepseek-v4-flash，即可使用最新版本。

Agent 能力显著增强，基准测试结果远超 V4-Pro-Preview：

  • Terminal Bench 2.1：82.7
  • NL2Repo：54.2
  • Cybergym：76.7
  • DeepSWE：54.4
  • Toolathlon verified：70.3
  • Agent Last Exam：25.2
  • Automation Bench（Public）：25.1
  • DSBench-FullStack：68.7
  • DSBench-Hard：59.6

注 1：对于公开基准测试集中的 Code Agent 任务，正式版 DeepSeek-V4-Flash 使用 DeepSeek Harness 最简模式（即将发布）作为测试框架，采用 max effort 级别、topp=0.95 和 temperature=1.0
注 2：DSBench-FullStack 是内部全栈开发测试集，DSBench-Hard 是内部 Coding Agent 高难度问题测试集

正式版 V4-Flash 原生支持 Responses API 格式，并针对 Codex 进行了适配。具体配置请参阅文档。 DeepSeek-V4-Flash-0731 与 DeepSeek-V4-Flash-Preview 保持相同的模型架构和规模，仅进行了后训练。 注意：本次更新仅升级 DeepSeek-V4-Flash API。DeepSeek-V4-Pro API 以及 APP/WEB 模型保持不变。 DeepSeek-V4-Pro 正式版将很快发布。

日期：2026-04-24

DeepSeek-V4

DeepSeek API 现已支持 V4-Pro 和 V4-Flash，可通过 OpenAI ChatCompletions 接口和 Anthropic 接口调用。要访问新模型，base_url 保持不变，模型参数应设置为 deepseek-v4-pro 或 deepseek-v4-flash。

两个旧版 API 模型名称 deepseek-chat 和 deepseek-reasoner 将在三个月后（2026-07-24）停止服务。在当前过渡期内，这两个模型名称分别指向 deepseek-v4-flash 的非思考模式和思考模式。

更多详情请参阅 此文档。

日期：2025-12-01

DeepSeek-V3.2

deepseek-chat 和 deepseek-reasoner 均已升级至 DeepSeek-V3.2。

  • deepseek-chat 对应 DeepSeek-V3.2 的非思考模式
  • deepseek-reasoner 对应 DeepSeek-V3.2 的思考模式

DeepSeek-V3.2-Speciale

DeepSeek-V3.2-Speciale 通过临时端点提供服务：base_url="https://api.deepseek.com/v3.2_speciale_expires_on_20251215"。价格与 V3.2 相同，不支持工具调用，可用至 2025 年 12 月 15 日 15:59（UTC 时间）。

更多详情请参阅 此文档。

日期：2025-09-29

DeepSeek-V3.2-Exp

deepseek-chat 和 deepseek-reasoner 均已升级至 DeepSeek-V3.2-Exp。

  • deepseek-chat 对应 DeepSeek-V3.2-Exp 的非思考模式
  • deepseek-reasoner 对应 DeepSeek-V3.2-Exp 的思考模式

更多详情请参阅 此文档。

日期：2025-09-22

DeepSeek-V3.1-Terminus

deepseek-chat 和 deepseek-reasoner 均已升级至 DeepSeek-V3.1-Terminus。 deepseek-chat 对应 DeepSeek-V3.1-Terminus 的非思考模式，而 deepseek-reasoner 对应其思考模式。

本次更新在保持模型原有能力的同时，解决了用户反馈的问题，包括：

  • 语言一致性：减少中英混杂及偶发异常字符的情况；
  • Agent 能力：进一步优化 Code Agent 和 Search Agent 的性能。

日期：2025-08-21

DeepSeek-V3.1

deepseek-chat 和 deepseek-reasoner 均已升级至 DeepSeek-V3.1。 deepseek-chat 对应 DeepSeek-V3.1 的非思考模式，而 deepseek-reasoner 对应其思考模式。

  • DeepSeek-V3.1 的主要更新：
    • 混合推理架构：单一模型同时支持思考模式和非思考模式
    • 推理效率提升：与 DeepSeek-R1-0528 相比，DeepSeek-V3.1-Think 可在显著更短的时间内给出答案
    • Agent 能力增强：通过后训练优化，新模型在工具使用和智能体任务上实现重大提升
      • SWE-bench Verified：66.0
      • SWE-bench Multilingual：54.5
      • Terminal-bench：31.3

日期：2025-05-28

deepseek-reasoner

deepseek-reasoner 模型已升级至 DeepSeek-R1-0528：

  • 推理能力增强
    • 基准测试结果显著提升（Pass@1）
      • AIME 2025：70.0 → 87.5（+17.5）
      • GPQA：71.5 → 81.0（+9.5）
      • LCB_v6：63.5 → 73.3（+9.8）
      • Aider：57.0 → 71.6（+14.6）
    • 注意：与旧版 R1 相比，复杂推理任务可能消耗更多 token。
  • 前端开发优化
    • 生成的网页和游戏现具有更好的视觉美感。
  • 幻觉减少
    • 显著抑制了旧版 R1 中存在的幻觉问题。
  • 支持 JSON 输出和函数调用
    • 函数调用性能：
      • Tau-bench 得分：53.5（Airline）/ 63.9（Retail）

日期：2025-03-24

deepseek-chat

deepseek-chat 模型已升级至 DeepSeek-V3-0324：

  • 推理能力增强

    • 基准测试性能显著提升：
      • MMLU-Pro：75.9 → 81.2（+5.3）
      • GPQA：59.1 → 68.4（+9.3）
      • AIME：39.6 → 59.4（+19.8）
      • LiveCodeBench：39.2 → 49.2（+10.0）

  • 前端网页开发优化

    • 代码生成准确率提升
    • 网页和游戏前端更加美观

  • 中文写作能力升级

    • 风格和内容质量提升：
      • 与 R1 写作风格保持一致
      • 中长篇写作质量更高

  • 功能增强

    • 改进多轮交互式改写
    • 优化翻译质量和信件写作

  • 中文搜索能力提升

    • 增强报告分析请求，输出更加详尽

  • 函数调用改进

    • 提高 Function Calling 准确率，修复此前 V3 版本存在的问题

日期：2025-01-20

deepseek-reasoner

  • deepseek-reasoner 是我们的新模型 DeepSeek-R1。您可以通过指定 model='deepseek-reasoner' 调用 DeepSeek-V3。
  • 详情请参阅：[DeepSeek-R1 发布

English / 原文

Change Log

Date: 2026-10-17​

Opening up the in-house search API​

DeepSeek's web and API search capabilities are powered by our in-house DeepSeek search engine, which reliably handles hundreds of millions of search requests from around the world every day. DeepSeek Search is an AI-native general-purpose search engine that indexes hundreds of billions of web pages, supports multilingual search, and returns highly relevant, authoritative, and timely results in real time. It filters SEO spam, GEO manipulation, and low-quality AIGC through strict detection mechanisms, ensuring authoritative and trustworthy information for users and LLMs. For usage details, see the Search & Fetch guide.

DeepSeek-V4.1-Flash-1017 Update​

DeepSeek V4.1 Flash has been upgraded to DeepSeek-V4.1-Flash-1017. The official web and APP models have also been updated to the new version. The API calling method and pricing remain unchanged — simply use deepseek-flash to access the latest model.

Performance–cost Pareto improvement

Compared with the previous DeepSeek V4.1, the new model delivers substantial improvements across benchmarks:

  • HLE (w/o tools): 42.6 (44.0*)

  • Terminal-Bench 4.0: 42.3

  • DeepSWE v1.1: 77.0

  • ProgramBench: 29.1

  • NL2Repo-Bench: 68.2

  • Automation-Bench v1.0.6: 60.5

  • Agents' Last Exam: 35.4

  • Apex Bench v1.1: 72.1

  • Chartography (w/tools): 83.0

  • BabyVision (w/tools): 92.5

  • ZeroBench-main (w/tools): 51.0

  • Tested only on the pure-text subset of the HLE benchmark set. Terminal-Bench 4.0 was tested with the Claude Code framework, and DeepSWE with the Mini-SWE-V2 framework. ZeroBench-main was evaluated with Pass@5.

While improving performance, we also reduced the model's task runtime and output length, enabling users to complete tasks at lower cost and higher speed and achieving a new performance–cost Pareto optimum across multiple harnesses. We specifically optimized for DeepSeek Harness, Tencent CodeBuddy, WorkBuddy, and other platforms. Please upgrade to DeepSeek Harness 0.2.1-rc.1 or later for the best experience with the new model.

Date: 2026-09-10​

DeepSeek-V4.1-Flash Release​

Today, we officially release the DeepSeek-V4.1-Flash model. It is the smallest model in our new architecture family, with native multimodal visual understanding. The new architecture is designed for a higher capability ceiling, faster inference, higher throughput, and scaling to larger models.

  • GPQA Diamond: 90.9

  • HLE: 36.8 (39.1*)

  • Codeforces (Rating): 3471

  • MathArena Apex: 65.6

  • Terminal-Bench 2.1: 90.6

  • Terminal-Bench 3.0: 30.0

  • Terminal-Bench 4.0: 31.2

  • DeepSWE v1.1: 74.2

  • ProgramBench: 20.3

  • NL2Repo-Bench: 65.4

  • CyberGym: 88.1

  • SEC-Bench Pro: 62.8

  • ExploitGym: 15.3

  • HLE (w/tools): 63.9

  • Automation-Bench: 54.8

  • Agents' Last Exam: 31.8

  • Chartography (w/tools): 78.9

  • BabyVision (w/tools): 89.6

  • ZeroBench-main (w/tools): 49.0

  • Tested only on the pure-text subset of the HLE benchmark set.

API changes

DeepSeek V4.1 Flash is now available on the DeepSeek API with native multimodal support. Change the model name to deepseek-flash to call the latest V4.1 Flash model. The previous-generation models V4 Flash and V4 Flash Vision Exp have been retired; for compatibility, the model names deepseek-v4-flash and deepseek-v4-flash-vision-exp are temporarily routed to V4.1 Flash.

In response to user demand, we have decided to continue providing API services for DeepSeek V4 Pro after September 14, 2026, with the billing method remaining unchanged. We will provide further notice should there be any changes. Thank you for your understanding and support!

API pricing adjustment

With the release of DeepSeek-V4.1-Flash, API prices have been reduced accordingly. For details, please refer to Models & Pricing.

For more details, please refer to this documentation.

Date: 2026-08-21​

DeepSeek-V4-Flash-Vision-Exp Release​

Today, the new multimodal vision understanding model DeepSeek-V4-Flash-Vision-Exp is now available on the DeepSeek API platform. This is an experimental model that can be accessed by setting model='deepseek-v4-flash-vision-exp'.

  • Terminal Bench 2.1: 83.9

  • NL2Repo: 57.7

  • DeepSWE: 59.3

  • DSBench-Hard: 63.6

  • AutomationBench (Public): 25.7

  • ApexBench (Pass@1): 36.5

  • Agents' Last Exam: 27.3

  • Chartography: 64.3

  • ZeroBench (Pass@5): 35.0

  • For the Code Agent text tasks in the public benchmark sets, the DeepSeek family models were tested using the DeepSeek Harness minimal mode as the framework, with the max effort level, topp=0.95, and temperature=1.0; in the ApexBench and Agents' Last Exam evaluations, the text model DeepSeek-V4-Flash ignores the multimodal elements within them.

In terms of pure-text capabilities (agent, reasoning, world knowledge, etc.), DeepSeek-V4-Flash-Vision-Exp is on par with the official DeepSeek-V4-Flash.

On agent benchmarks that require visual understanding, DeepSeek-V4-Flash-Vision-Exp delivers a significant leap over DeepSeek-V4-Flash, bringing its multimodal agent capabilities close to Opus-4.8.

For usage details, please refer to the Vision guide.

Date: 2026-08-13​

DeepSeek-V4-Pro Update​

The GA release of DeepSeek-V4-Pro has been rolled out on the APP, Web, and API. The API calling method remains unchanged — simply set the model name to deepseek-v4-pro to use the latest version.

Significantly enhanced Agent capabilities

The GA version of DeepSeek V4 Pro greatly enhances agent capabilities, with particularly significant performance improvements in production environments.

  • HLE (wo / w tools): 42.7/60.0
  • Terminal Bench 2.1: 87.9
  • NL2Repo: 61.5
  • Cybergym: 83.3
  • DeepSWE: 62.7
  • Toolathlon-Verified: 74.1
  • Agents' Last Exam: 25.7
  • AutomationBench (Public): 31.8
  • DSBench-FullStack: 71.1
  • DSBench-Hard: 67.2

Native support for the Responses API

The DeepSeek API now natively supports the OpenAI Responses API format and is specifically adapted for Codex. Users can refer to the official documentation and complete the Codex configuration with a one-click configuration script.

More flexible thinking effort control

The thinking modes of V4-Pro and V4-Flash now support three thinking effort levels: low / high / max. In real-world usage, users can flexibly choose based on task complexity: use low for simple tasks, high for daily Agent tasks, and max for more complex scenarios. For setup instructions, please refer to the official API documentation: Thinking Mode.

API Pricing Adjustment

With the official release of the DeepSeek V4 model family, we will update and adjust API pricing. To allocate resources more reasonably, we will adopt peak/off-peak pricing, with off-peak prices set at half of the peak-hour prices, encouraging users to schedule their tasks based on actual usage. The new prices will take effect at 16:00 (UTC Time) on August 16, 2026.

For more details, please refer to this documentation.

Date: 2026-07-31​

DeepSeek-V4-Flash Update​

The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version.

Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview:

  • Terminal Bench 2.1: 82.7
  • NL2Repo: 54.2
  • Cybergym: 76.7
  • DeepSWE: 54.4
  • Toolathlon verified: 70.3
  • Agent Last Exam: 25.2
  • Automation Bench (Public): 25.1
  • DSBench-FullStack: 68.7
  • DSBench-Hard: 59.6

Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0
Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set

The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to thedocumentation. DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This update only upgrades the DeepSeek-V4-Flash API. The DeepSeek-V4-Pro API and the APP/WEB models are unchanged. The official release of DeepSeek-V4-Pro will follow soon.

Date: 2026-04-24​

DeepSeek-V4​

The DeepSeek API now supports V4-Pro and V4-Flash, available via both the OpenAI ChatCompletions interface and the Anthropic interface. To access the new models, the base_url remains unchanged, and the model parameter should be set to deepseek-v4-pro or deepseek-v4-flash.

The two legacy API model names, deepseek-chat and deepseek-reasoner, will be discontinued in three months (2026-07-24). During the current period, these two model names point to the non-thinking mode and thinking mode of deepseek-v4-flash, respectively.

For more details, please refer to this documentation.

Date: 2025-12-01​

DeepSeek-V3.2​

Both deepseek-chat and deepseek-reasoner have been upgraded to DeepSeek-V3.2.

  • deepseek-chat corresponds to DeepSeek-V3.2's non-thinking mode
  • deepseek-reasoner corresponds to DeepSeek-V3.2's thinking mode

DeepSeek-V3.2-Speciale​

DeepSeek-V3.2-Speciale is served via a temporary endpoint: base_url="https://api.deepseek.com/v3.2_speciale_expires_on_20251215". Same pricing as V3.2, no tool calls, available until Dec 15th, 2025, 15:59 (UTC Time).

For more details, please refer to this documentation.

Date: 2025-09-29​

DeepSeek-V3.2-Exp​

Both deepseek-chat and deepseek-reasoner have been upgraded to DeepSeek-V3.2-Exp.

  • deepseek-chat corresponds to DeepSeek-V3.2-Exp's non-thinking mode
  • deepseek-reasoner corresponds to DeepSeek-V3.2-Exp's thinking mode

For more details, please refer to this documentation.

Date: 2025-09-22​

DeepSeek-V3.1-Terminus​

Bothdeepseek-chat and deepseek-reasoner have been upgraded to DeepSeek-V3.1-Terminus. deepseek-chat corresponds to DeepSeek-V3.1-Terminus's non-thinking mode , while deepseek-reasoner corresponds to its thinking mode.

This update maintains the model's original capabilities while addressing issues reported by users, including:

  • Language consistency: Reduced occurrences of Chinese-English mixing and occasional abnormal characters;
  • Agent capabilities: Further optimized the performance of the Code Agent and Search Agent.

Date: 2025-08-21​

DeepSeek-V3.1​

Bothdeepseek-chat and deepseek-reasoner have been upgraded to DeepSeek-V3.1. deepseek-chat corresponds to DeepSeek-V3.1's non-thinking mode , while deepseek-reasoner corresponds to its thinking mode.

  • Key updates in DeepSeek-V3.1:
    • Hybrid reasoning architecture : A single model supports both thinking mode and non-thinking mode
    • Improved reasoning efficiency : Compared to DeepSeek-R1-0528, DeepSeek-V3.1-Think provides answers in significantly less time
    • Enhanced agent capabilities : With post-training optimization, the new model achieves major improvements in tool usage and intelligent agent tasks
      • SWE-bench Verified: 66.0
      • SWE-bench Multilingual: 54.5
      • Terminal-bench: 31.3

Date: 2025-05-28​

deepseek-reasoner​

deepseek-reasoner Model Upgraded to DeepSeek-R1-0528:

  • Enhanced Reasoning Capabilities
    • Significant benchmark improvements (Pass@1)
      • AIME 2025: 70.0 → 87.5 (+17.5)
      • GPQA: 71.5 → 81.0 (+9.5)
      • LCB_v6: 63.5 → 73.3 (+9.8)
      • Aider: 57.0 → 71.6 (+14.6)
    • Note: Complex reasoning tasks may consume more tokens compared to legacy R1 version.
  • Optimized Front-end Development
    • Generated web pages and games now feature improved aesthetics.
  • Reduced Hallucinations
    • Significantly suppressed hallucination issues present in legacy R1 version.
  • JSON Output & Function Calling Support
    • Function call performance:
      • Tau-bench score: 53.5 (Airline) / 63.9 (Retail)

Date: 2025-03-24​

deepseek-chat​

deepseek-chat Model Upgraded to DeepSeek-V3-0324:

  • Enhanced Reasoning Capabilities

    • Significant improvements in benchmark performance:
      • MMLU-Pro: 75.9 → 81.2 (+5.3)
      • GPQA: 59.1 → 68.4 (+9.3)
      • AIME: 39.6 → 59.4 (+19.8)
      • LiveCodeBench: 39.2 → 49.2 (+10.0)

  • Optimized Front-End Web Development

    • Improved accuracy in code generation
    • More aesthetically pleasing web pages and game front-ends

  • Upgraded Chinese Writing Proficiency

    • Enhanced style and content quality:
      • Aligned with the R1 writing style
      • Better quality in medium-to-long-form writing

  • Feature Enhancements

    • Improved multi-turn interactive rewriting
    • Optimized translation quality and letter writing

  • Improved Chinese Search Capabilities

    • Enhanced report analysis requests with more detailed outputs

  • Function Calling Improvements

    • Increased accuracy in Function Calling, fixing issues from previous V3 versions

Date: 2025-01-20​

deepseek-reasoner​

  • deepseek-reasoner is our new model DeepSeek-R1. You can invoke DeepSeek-V3 by specifying model='deepseek-reasoner'.
  • For details, please refer to: DeepSeek-R1 Release
  • For guides, please refer to: Thinking Mode

Date: 2024-12-26​

deepseek-chat​

  • The deepseek-chat model has been upgraded to DeepSeek-V3. The API remains unchanged. You can invoke DeepSeek-V3 by specifying model='deepseek-chat'.
  • For details, please refer to: introducing DeepSeek-V3

Date: 2024-12-10​

deepseek-chat​

The deepseek-chat model has been upgraded to DeepSeek-V2.5-1210 , with improvements across various capabilities. Relevant benchmarking results include:

  • Mathematical: Performance on the MATH-500 benchmark has improved from 74.8% to 82.8% .
  • Coding: Accuracy on the LiveCodebench (08.01 - 12.01) benchmark has increased from 29.2% to 34.38% .
  • Writing and Reasoning: Corresponding improvements have been observed in internal test datasets.

Additionally, the new version of the model has optimized the user experience for file upload and webpage summarization functionalities.

Date: 2024-09-05​

deepseek-coder & deepseek-chat Upgraded to DeepSeek V2.5 Model​

The DeepSeek V2 Chat and DeepSeek Coder V2 models have been merged and upgraded into the new model, DeepSeek V2.5.

For backward compatibility, API users can access the new model through either deepseek-coder or deepseek-chat.

The new model significantly surpasses the previous versions in both general capabilities and code abilities.

The new model better aligns with human preferences and has been optimized in various areas such as writing tasks and instruction following:

  • ArenaHard win rate improved from 68.3% to 76.3%
  • AlpacaEval 2.0 LC win rate increased from 46.61% to 50.52%
  • MT-Bench score rose from 8.84 to 9.02
  • AlignBench score increased from 7.88 to 8.04

The new model has further enhanced its code generation capabilities based on the original Coder model, optimized for common programming application scenarios, and achieved the following results on the standard test set:

  • HumanEval: 89%
  • LiveCodeBench (January-September): 41%

Date: 2024-08-02​

API Launches Context Caching on Disk Technology​

The DeepSeek API has innovatively adopted hard disk caching, reducing prices by another order of magnitude.

For more details on the update, please refer to the documentation Context Caching is Available 2024/08/02.

Date: 2024-07-25​

New API Features​

  • Update API /chat/completions
    • JSON Mode
    • Function Calling
    • Chat Prefix Completion（Beta）
    • 8K max_tokens（Beta）
  • New API /completions

For more details, please check the documentation New API Features 2024/07/25

Date: 2024-07-24​

deepseek-coder​

The deepseek-coder model has been upgraded to DeepSeek-Coder-V2-0724.

Date: 2024-06-28​

deepseek-chat​

The deepseek-chat model has been upgraded to DeepSeek-V2-0628.

Model's reasoning capabilities have improved, as shown in relevant benchmarks:

  • Coding: HumanEval Pass@1 79.88% -> 84.76%
  • Mathematics: MATH ACC@1 55.02% -> 71.02%
  • Reasoning: BBH 78.56% -> 83.40%

In the Arena-Hard evaluation, the win rate against GPT-4-0314 increased from 41.6% to 68.3%.

The model's role-playing capabilities have significantly enhanced, allowing it to act as different characters as requested during conversations.

Date: 2024-06-14​

deepseek-coder​

The deepseek-coder model has been upgraded to DeepSeek-Coder-V2-0614, significantly enhancing its coding capabilities. It has reached the level of GPT-4-Turbo-0409 in code generation, code understanding, code debugging, and code completion. Additionally, it possesses excellent mathematical and reasoning abilities, and its general capabilities are on par with DeepSeek-V2-0517.

Date: 2024-05-17​

deepseek-chat​

The deepseek-chat model has been upgraded to DeepSeek-V2-0517. The model has seen a significant improvement in following instructions, with the IFEval Benchmark Prompt-Level accuracy jumping from 63.9% to 77.6%. Additionally, on API end, we have optimized model ability to follow instruction filled in the ``system" part. This optimization has significantly elevated the user experience across a variety of tasks, including immersive translation, Retrieval-Augmented Generation (RAG), and more.

The model's accuracy in outputting JSON format has been enhanced. In our internal test set, the JSON parsing rate increased from 78% to 85%. By introducing appropriate regular expressions, the JSON parsing rate was further improved to 97%.