DeepSeek
@deepseek_aiOct 17
🚀 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

Oct 17, 2026 · 8:35 AM UTC

1,611
4,127
2,821
37,024
7,831,723
RelevantRecentLikes
DeepSeek
@deepseek_aiOct 17
🎮 Build complete projects from scratch DeepSeek-V4.1-Flash-1017 uses reinforcement learning with self-feedback to improve end-to-end project development—from game creation to 3D modeling. See it in action 👇 2/4
59
100
58
2,379
521,020
DeepSeek
@deepseek_aiOct 17
🔍 Search & Fetch APIs are now open to everyone! Powered by our in-house search engine, serving hundreds of millions of requests daily worldwide. 💰 Each endpoint: $0.75 per 1,000 calls. 📖 Docs: api-docs.deepseek.com/guides… 📌 Search & Fetch calls via server-side tools in DeepSeek Harness, Claude Code, and Codex are currently free for DeepSeek API users. Billing starts Oct 31, 2026, at 16:00 UTC. 3/4

Search & Fetch | DeepSeek API Docs

The Search API and the Fetch API give agents and RAG applications access to live web content:

api-docs.deepseek.com
89
248
171
3,482
955,027
DeepSeek
@deepseek_aiOct 17
🔓 Open-source release Get DeepSeek-V4.1-Flash-1017: 🤗 Hugging Face: huggingface.co/deepseek-ai/D… 🔗 ModelScope: modelscope.cn/models/deepsee… 4/4

deepseek-ai/DeepSeek-V4.1-Flash-1017 · Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

huggingface.co
48
224
48
2,919
336,909
@deepseek_ai @DeepSeekHarness
🤖 Made with AI
9
18
1
2,258
130,068
Whale 🐋 is back babyyyy
4
13
1
1,287
71,029
Deeeeeeeeeeepseeeeeeeeek
1
1
183
51,382
i love you deepseek please marry me
18
10
1
938
96,999
实测了！DeepSeek-V4.1-Flash-1017 做动效的水平非常牛，基本上仅次于Claude模型的水平。 用GPT-6暂时也没做出这么好的效果。 当然，这是搭配了我新做的动效skill做到的，预计下周发布，敬请期待。
31
19
5
846
110,019
Sam after see DeepSeek beat GPT-5.6-Sol.
3
6
1
395
37,308
Evan Kirstel #B2B #TechFluencer
@EvanKirstelOct 17
@deepseek_ai @DeepSeekHarness
4
1
55
16,330
The level of detail in the technical report is so amazing
5
3
145
33,737
感受我DeepSeek的压制力吧，达里奥。
10
297
27,387
@deepseek_ai @DeepSeekHarness
4
3,741
@deepseek_ai @DeepSeekHarness
3
3,595
Your naming conventions are confusing. Just name it 4.2
20
557
82,044
You guys are honestly the GOAT of LLMs for the common man! We love you 😍
1
107
32,289
My world is better because of 4.1-Flash. And I'm creating a better world because of this, DSH is my daily driver.
1
76
24,574
open model better than 5.6 sol nice
3
471
59,622
DeepSeek新模型做视频到底行不行？？我看X上还挺多不同评价的。👇这是DeepSeek-V4.1-flash花了225分钟17.2元，做出来的鲸鱼穿越艺术史的动画。我觉得基本上能达到Opus5.5 80%左右的水准，略好于GPT-6 Astra。当然，这种好是在合适的harness和skill加持下实现的。这次做视频用的是我已经开源的huashu-art-motion：github.com/alchaincyf/huashu…
2
1
1
30
9,125
shorter outputs i've been asking for that in every prompt for months usually the answer is four paragraphs promising to keep it short
4
6,152
Mega thanks for this release 🙏
1
56
27,459
You don't understand how version numbers work. 🤦‍♂️😅 Just call it 4.2!🤓
4
1,401
太恐怖了！从发布到现在我拿新版DeepSeek V4.1 flash测了10多个复杂的agent任务，以及花了接近5亿token。 它的缓存命中率达到了恐怖的99.41%...好夸张，DeepSeek在缓存降成本能力上一骑绝尘了
3
3
155
30,653
Congratulations 🎉 open source 🔥
46
28,415
@deepseek_ai @DeepSeekHarness
1
12
9,590
Lincoln 🇿🇦
@PresidentlinOct 17
Nice minor update Can't wait for Pro As always oblig How high will your ceiling go?!!
1
126
31,517
impressive! congrats on the release yall!
3
1
154
35,751
If this is what updated flash is, what in the world is next pro going to look like ?
13
271
56,574
opening the search/fetch stack alongside the model is a smart move. weights alone only get you so far now. retrieval + harness quality is becoming a huge chunk of whether the agent actually feels good to use.
8
9,424
you're so sneaky: opus 5 wasn't usable
1,727
amazing congrats!
31
25,709
why are we comparing this against prev gen models though? sol 6.1, opus 5.5...
1
4
6,665
Thanks for the open weights! Used to spar with GLM-5.3-flash but now better than GLM-5.3 at nearly everything. Need to test the cyber capabilities now
39
20,552
Huge win stronger faster and more efficient all at once thats the perfect Pareto improvement
75
Still opus 5 is ahead even here.
863
又是梁子化身梁圣的一天，虽然版本号都没变，只改了版本日期，但相比年初的V4版本，这至少该是V4.5 Pro级别的模型了。
7
5
210
55,918
Load more replies