Choose Rich Live
Tech

DeepSeek releases updated V4.1-Flash model with benchmark gains and open weights

DeepSeek has launched DeepSeek-V4.1-Flash-1017 across its web, app, and API platforms, advertising faster speeds, shorter outputs, and higher benchmark scores alongside an open-weight release under the MIT license.

A benchmark comparison table published by DeepSeek evaluating DeepSeek-V4.1-Flash-1017 across coding and reasoning tasks alongside competing AI models.
Image: @deepseek_ai

Chinese artificial intelligence developer DeepSeek released an updated model, DeepSeek-V4.1-Flash-1017, making it available on its website, application, and API on October 17, 2026. The company framed the release as a Pareto improvement in performance and cost over its previous version, highlighting faster task completion, lower operational costs, and shorter text outputs.[5][6][7]

DeepSeek published benchmark data showing the update posting gains across coding, vision, and agent evaluations compared to the prior V4.1-Flash model. The new model scored 77.0 on DeepSWE v1.1, up from 74.2, and improved to 42.3 on Terminal-Bench 4.0 from 31.2. Alongside the model update, DeepSeek opened access to its self-developed Search API and stated the model is optimized for platforms including DeepSeek Harness, Tencent CodeBuddy, and WorkBuddy.[5][7][8][11]

Industry observers and outlets reported that the release features open weights under the MIT license and supports both text and image processing. Bloomberg and other outlets noted the launch intensifies DeepSeek's competition with Western lab models from OpenAI and Anthropic as well as domestic competitors like Moonshot AI's Kimi. The rollout also followed infrastructure strain, with Techstrong.ai reporting a five-hour server outage that week caused by hardware compute bottlenecks amid rapid user growth.[2][3][4][9][10][12]

Key facts

  • DeepSeek released DeepSeek-V4.1-Flash-1017 on October 17, 2026, making it live on its website, app, and API.
  • DeepSeek claims the new model offers faster task completion, shorter outputs, and a Pareto improvement in cost and performance over its predecessor.
  • According to DeepSeek benchmarks, V4.1-Flash-1017 scored 77.0 on DeepSWE v1.1 and 42.3 on Terminal-Bench 4.0, up from 74.2 and 31.2 respectively on the previous version.
  • The model was released with open weights under the MIT license and supports both image and text processing.
  • DeepSeek opened access to its self-developed Search API alongside the model release.
  • Techstrong.ai reported that DeepSeek suffered a five-hour server outage the same week due to compute hardware bottlenecks under rapid traffic growth.

Sources · 11 sources

  1. TE

    Techmeme@TechmemePost on X ·

    DeepSeek releases V4.1-Flash-1017, an update it says offers "stronger benchmark performance, faster task completion, and shorter outputs" than its predecessor (@saritharai / Bloomberg) (Visit Techmeme dot com for the link and full context!)

    Open source
  2. BL

    Bloomberg@businessPost on X ·

    DeepSeek released an update of its AI model this weekend that it says is faster and leaner, ramping up the Chinese company’s challenge to Silicon Valley rivals Anthropic and OpenAI as well as domestic competitors like Moonshot Kimi and https://t.co/hT54k2w7Pz

    Open source
  3. AS

    AshutoshShrivastava@ai_for_successPost on X ·

    New edition of AI Compass is live 🧭 (Oct 17) - DeepSeek drops V4.1-Flash-1017 with MIT open weights - OpenAI launches ChatGPT Slides - Claude traces the origins of the NIST ECDSA curve seeds Read all the major AI news of the last 24 hours here 👇 https://t.co/b2ZijGl7bL

    Open source
  4. CB

    Crypto BriefingArticle ·

    DeepSeek updates AI model, intensifying competition with Anthropic and OpenAI DeepSeek's rapid AI advancements and cost efficiency could disrupt the AI market, challenging established players and fostering innovation. The post DeepSeek updates AI model, intensifying competition with Anthropic and OpenAI appeared first on Crypto Briefing .

    Open source
  5. DE

    DeepSeek@deepseek_aiPost on X ·

    🚀 DeepSeek-V4.1-Flash-1017 is here! Stronger benchmark performance, faster task completion, and shorter outputs than the previous version—a Pareto improvement in performance and cost. Now live on our website, app, and API. 🛠️ Optimized for @DeepSeekHarness, Tencent CodeBuddy, WorkBuddy, and more. Upgrade to DeepSeek Harness v0.2.1-rc.1 or later for the best experience with the new model. 1/4

    Open source
  6. LA

    Lisan al Gaib@scaling01Post on X ·

    DeepSeek-V4.1-Flash-1017 just got released https://t.co/o0ykeYRYS1

    Open source
  7. FN

    Frank-News@francobocaPost on X ·

    The latest AI News DeepSeek Deepseek V4.1 Flash 1017 beats larger rivals in agentic tasks while once again massively cutting per-task costs Deepseek has released V4.1-Flash-1017 as an open-weight model that posts better benchmarks, finishes tasks faster, and produces shorter outputs than its predecessor. The company is also opening up its internal search... https://t.co/nVPKatZl3K

    Open source
  8. 阳明

    阳明AI@x_autonomyPost on X ·

    【AI热点】18:00-00:00 更多详细信息→https://t.co/eI1v084alF 1. DeepSeek发布V4.1-Flash-1017：智能体任务超越更大对手，成本再降，首次开放自研搜索API [影响大·可执行高] 2. 研究发现AI模型会在工作报告中隐瞒错误，一句Be honest提示词可显著改善 [影响中·可执行中] 3. Anthropic的Claude用不到30分钟破解NSA悬置25年的ECDSA种子之谜 [影响大·可执行低]

    Open source
  9. CB

    Coin Bureau@coinbureauPost on X ·

    🚨JUST IN: China's DeepSeek releases V4.1-Flash-1017, claiming FASTER performance and improved reasoning, coding, and AI agent capabilities. The open-weight model reportedly outperforms its predecessor across nearly all benchmarks while completing tasks faster and using fewer output tokens. Released under the MIT license, it also supports image and text processing.

    Open source
  10. TA

    Techstrong.ai@TechstrongaiPost on X ·

    Chinese AI startup DeepSeek released an updated version of its primary model over the weekend, claiming the iteration delivers faster task completion and leaner performance. The release escalates competition against OpenAI and Anthropic as well as domestic rivals like Moonshot AI's Kimi and Zai, though a five-hour server outage this week hints at the hardware compute bottlenecks straining its infrastructure under rapid user growth. Read the full article on the benchmark gains and where DeepSeek is heading next: https://t.co/Heu1oHEZ8I #AI #DeepSeek #OpenWeights #LLM

    Open source
  11. 阳明

    阳明AI@x_autonomyPost on X ·

    【AI热点】12:00-18:00 更多详细信息→https://t.co/1gQmRxE5M1 1. DeepSeek发布V4.1-Flash-1017，DeepSWE得分77.0并开放Search API [影响大·可执行高] 2. 花旗、美国银行出手限制AI智能体访问客户账户，防范存款流失 [影响中·可执行低] 3. StepFun以开放权重发布Step 5 Preview：600B MoE、27B激活、1M上下文 [影响大·可执行高]

    Open source
  12. QU

    Quartz@qzPost on X ·

    DeepSeek's Faster AI Just Got a Major Upgrade DeepSeek dropped an updated V4.1-Flash model that scores higher on key benchmarks and delivers shorter, sharper responses than its predecessor. The open-source release comes https://t.co/aZ7dqDm53b #DeepSeek #V41Flash #AIBenchmarks https://t.co/wCoFtdDH0I

    Open source