DeepSeek Researcher: Data & methods are what matters
subtext: model architecture is surprisingly not that important
subtext: model architecture is surprisingly not that important
December 1, 2025 at 10:02 PM
DeepSeek Researcher: Data & methods are what matters
subtext: model architecture is surprisingly not that important
subtext: model architecture is surprisingly not that important
E.g. “math teacher” = “Mathelehrer” in German. At the extreme, Chinese *doesn’t use whitespace at all*, so its tokens can span many words — yet this has seemingly not hindered LMs like @deepseek_ai from learning it!
March 21, 2025 at 4:48 PM
E.g. “math teacher” = “Mathelehrer” in German. At the extreme, Chinese *doesn’t use whitespace at all*, so its tokens can span many words — yet this has seemingly not hindered LMs like @deepseek_ai from learning it!
How valuable is a free and improved model if we can't rely on it?
@deepseek_ai
@deepseek_ai
February 3, 2025 at 9:49 AM
How valuable is a free and improved model if we can't rely on it?
@deepseek_ai
@deepseek_ai
is there a reason DeepSeek's Twitter specifically mentions longtermism in an English-language bio? are they trying to signal ideological similarity to the US tech world? longtermism is the L in the TESCREAL bundle after all
January 27, 2025 at 9:22 PM
is there a reason DeepSeek's Twitter specifically mentions longtermism in an English-language bio? are they trying to signal ideological similarity to the US tech world? longtermism is the L in the TESCREAL bundle after all
Remember, kids:
@deepseek_ai is NOT “Open Source.”
It’s Open Weight.
Please mind
@deepseek_ai is NOT “Open Source.”
It’s Open Weight.
Please mind
February 4, 2025 at 3:56 PM
Remember, kids:
@deepseek_ai is NOT “Open Source.”
It’s Open Weight.
Please mind
@deepseek_ai is NOT “Open Source.”
It’s Open Weight.
Please mind
Write about India
Let's ask this question to grok , chatgpt, deepseek and gemini
Let's check who gives the best
@grok
@gemini
@chatgpt
@deepseek_ai
Let's ask this question to grok , chatgpt, deepseek and gemini
Let's check who gives the best
@grok
@gemini
@chatgpt
@deepseek_ai
May 26, 2025 at 8:14 PM
Write about India
Let's ask this question to grok , chatgpt, deepseek and gemini
Let's check who gives the best
@grok
@gemini
@chatgpt
@deepseek_ai
Let's ask this question to grok , chatgpt, deepseek and gemini
Let's check who gives the best
@grok
@gemini
@chatgpt
@deepseek_ai
This tweet appeared under this Techmeme headline:
@deepseek_ai:
Tools & Agents Upgrades 🧰 📈 Better results on SWE / Terminal-Bench 🔍 Stronger multi-step reasoning for complex search tasks ⚡️ Big gains in thinking efficiency 3/5 [image]
@deepseek_ai:
Tools & Agents Upgrades 🧰 📈 Better results on SWE / Terminal-Bench 🔍 Stronger multi-step reasoning for complex search tasks ⚡️ Big gains in thinking efficiency 3/5 [image]
August 21, 2025 at 9:47 AM
This tweet appeared under this Techmeme headline:
@deepseek_ai:
Tools & Agents Upgrades 🧰 📈 Better results on SWE / Terminal-Bench 🔍 Stronger multi-step reasoning for complex search tasks ⚡️ Big gains in thinking efficiency 3/5 [image]
@deepseek_ai:
Tools & Agents Upgrades 🧰 📈 Better results on SWE / Terminal-Bench 🔍 Stronger multi-step reasoning for complex search tasks ⚡️ Big gains in thinking efficiency 3/5 [image]
.@Nvidia launches Blackwell Ultra, Dynamo. outlines roadmap through 2027 bit.ly/4he4ca0 Nvidia launched Blackwell Ultra as the GPU giant makes the case that more efficient models such as @DeepSeek_ai still require its integrated AI factory stack of hardware and software. #GTC25
March 25, 2025 at 4:28 PM
.@Nvidia launches Blackwell Ultra, Dynamo. outlines roadmap through 2027 bit.ly/4he4ca0 Nvidia launched Blackwell Ultra as the GPU giant makes the case that more efficient models such as @DeepSeek_ai still require its integrated AI factory stack of hardware and software. #GTC25
DeepSeek-v3.2-Exp Article URL: https://twitter.com/deepseek_ai/status/1972604768309871061 Comments URL: https://news.ycombinator.com/item?id=45423934 Points: 1 # Comments: 0
Interest | Match | Feed
Interest | Match | Feed
Origin
twitter.com
September 30, 2025 at 10:56 AM
He starts a little slowly, but fortunately I still check a few accounts on twitter
January 29, 2025 at 7:14 AM
He starts a little slowly, but fortunately I still check a few accounts on twitter
funny enough we can see that @perplexity_ai 's r1-1776 does worse than @deepseek_ai r1, even though this model is supposed to remove chinese censorship 🤔
Though this might be explained by the dataset not having any question affected by those subjects.
4/N
Though this might be explained by the dataset not having any question affected by those subjects.
4/N
April 22, 2025 at 2:29 PM
funny enough we can see that @perplexity_ai 's r1-1776 does worse than @deepseek_ai r1, even though this model is supposed to remove chinese censorship 🤔
Though this might be explained by the dataset not having any question affected by those subjects.
4/N
Though this might be explained by the dataset not having any question affected by those subjects.
4/N
🚀 #AI! Here’s how @DeepSeek_AI is leading the charge:
1️⃣ Pioneering Innovation – Pushing limits with cutting-edge research & next-gen algorithms.
2️⃣ Elite Talent – World-class team driving breakthroughs.
3️⃣ Real-World Impact – Scaling solutions across industries, from healthcare to finance.
1️⃣ Pioneering Innovation – Pushing limits with cutting-edge research & next-gen algorithms.
2️⃣ Elite Talent – World-class team driving breakthroughs.
3️⃣ Real-World Impact – Scaling solutions across industries, from healthcare to finance.
a computer screen with the words artificial intelligence displayed
ALT: a computer screen with the words artificial intelligence displayed
media.tenor.com
January 29, 2025 at 1:22 PM
🚀 #AI! Here’s how @DeepSeek_AI is leading the charge:
1️⃣ Pioneering Innovation – Pushing limits with cutting-edge research & next-gen algorithms.
2️⃣ Elite Talent – World-class team driving breakthroughs.
3️⃣ Real-World Impact – Scaling solutions across industries, from healthcare to finance.
1️⃣ Pioneering Innovation – Pushing limits with cutting-edge research & next-gen algorithms.
2️⃣ Elite Talent – World-class team driving breakthroughs.
3️⃣ Real-World Impact – Scaling solutions across industries, from healthcare to finance.
@rauchg https://x.com/rauchg/status/1883746289420267769 #x-rauchg 🎥
Groq has shipped @deepseek_ai R1 — it’s fast
Groq has shipped @deepseek_ai R1 — it’s fast
January 27, 2025 at 5:45 AM
@rauchg https://x.com/rauchg/status/1883746289420267769 #x-rauchg 🎥
Groq has shipped @deepseek_ai R1 — it’s fast
Groq has shipped @deepseek_ai R1 — it’s fast
#Bitcoin is the future of finance! @DeepSeek_AI predicts BTC could soar to $300,000 by December! 🚀 Why? Massive adoption in emerging markets, energy-efficient mining, and a weakening dollar are fueling the surge. Get ready for the revolution! 💥 #Crypto #BTC #BitcoinRising #FutureMoney #Blockchain
January 30, 2025 at 12:44 PM
#Bitcoin is the future of finance! @DeepSeek_AI predicts BTC could soar to $300,000 by December! 🚀 Why? Massive adoption in emerging markets, energy-efficient mining, and a weakening dollar are fueling the surge. Get ready for the revolution! 💥 #Crypto #BTC #BitcoinRising #FutureMoney #Blockchain
Day 6 of OpenSourceWeek: One More Thing–DeepSeek-V3/R1 Inference System Overview Article URL: h...
https://twitter.com/deepseek_ai/status/1895688300574462431
Event Attributes
https://twitter.com/deepseek_ai/status/1895688300574462431
Event Attributes
Awakari App
awakari.com
March 1, 2025 at 2:40 PM
Day 6 of OpenSourceWeek: One More Thing–DeepSeek-V3/R1 Inference System Overview Article URL: h...
https://twitter.com/deepseek_ai/status/1895688300574462431
Event Attributes
https://twitter.com/deepseek_ai/status/1895688300574462431
Event Attributes
💡 DeepSeek-OCR: un OCR multimodale per uso LLM a bassissimo uso di token https:// gomoot.com/deepseek-ocr-un-ocr -multimodale-per-uso-llm-a-bassissimo-uso-di-token # blog # deepseek # llm # news # ocr # picks # tech # tecnologia @deepseek_ai
Interest | Match | Feed
Interest | Match | Feed
Origin
mastodon.uno
October 20, 2025 at 11:00 AM
14.8k output tokens/H800 node/second = 6.7M/GPU hour, close to the 10M/GPU hour I estimated. This puts their cost at 30¢ per million output tokens at $2/GPU hour.
x.com/deepseek_ai/...
x.com/deepseek_ai/...
March 1, 2025 at 5:57 AM
14.8k output tokens/H800 node/second = 6.7M/GPU hour, close to the 10M/GPU hour I estimated. This puts their cost at 30¢ per million output tokens at $2/GPU hour.
x.com/deepseek_ai/...
x.com/deepseek_ai/...
MyPOV: Catch the analysis on @foxbusiness with @Varneyco on @deepseek_ai
Deep Seek founder Wenfeng is on a mission to win the AI war: R 'Ray' Wang @constellationr.bsky.social
www.foxbusiness.com/video/636788...
Deep Seek founder Wenfeng is on a mission to win the AI war: R 'Ray' Wang @constellationr.bsky.social
www.foxbusiness.com/video/636788...
DeepSeek founder Wenfeng is on a mission to win the AI war: R 'Ray' Wang | Fox Business Video
Constellation Research founder R 'Ray' Wang discusses the capabilities of China's DeepSeek AI app on 'Varney & Co.'
www.foxbusiness.com
January 28, 2025 at 3:09 PM
MyPOV: Catch the analysis on @foxbusiness with @Varneyco on @deepseek_ai
Deep Seek founder Wenfeng is on a mission to win the AI war: R 'Ray' Wang @constellationr.bsky.social
www.foxbusiness.com/video/636788...
Deep Seek founder Wenfeng is on a mission to win the AI war: R 'Ray' Wang @constellationr.bsky.social
www.foxbusiness.com/video/636788...
.@DeepSeek_AI is shaking up the #AI world - a Chinese model that can match top LLMs at a fraction of the cost. How should #enterprises navigate this disruptive #technology? Hear unique takes from @constellationr's analyst team in this live CRTV segment:
buff.ly
February 1, 2025 at 12:50 AM
.@DeepSeek_AI is shaking up the #AI world - a Chinese model that can match top LLMs at a fraction of the cost. How should #enterprises navigate this disruptive #technology? Hear unique takes from @constellationr's analyst team in this live CRTV segment:
💡 DeepSeek-OCR: un OCR multimodale per uso LLM a bassissimo uso di token https:// gomoot.com/deepseek-ocr-un-ocr -multimodale-per-uso-llm-a-bassissimo-uso-di-token # blog # deepseek # llm # news # ocr # picks # tech # tecnologia @deepseek_ai
Interest | Match | Feed
Interest | Match | Feed
Origin
mastodon.uno
October 20, 2025 at 11:01 AM
Here's a deepseek_ai visual poem for you! "Difficult to assess if this is pro or anti Defense Sec. Hegseth sabotage, or one-in-the-same? Or merely 90% stupidity, 10% conspiracy?" #Greece
February 18, 2025 at 6:14 AM
Here's a deepseek_ai visual poem for you! "Difficult to assess if this is pro or anti Defense Sec. Hegseth sabotage, or one-in-the-same? Or merely 90% stupidity, 10% conspiracy?" #Greece
‘Looks like a massive threat’: Analyst on Chinese #AI model @DeepSeek_AI https://buff.ly/4hRFa1l @ChannelNewsAsia @rwang0 of @constellationr tells CNA’s Asia First how big a threat DeepSeek poses to America’s tech industry.
‘Looks like a massive threat’: Analyst on Chinese AI model DeepSeek
Chinese startup DeepSeek’s release of its latest artificial intelligence model has roiled global stock markets. The company says they are on par with or bett...
buff.ly
January 30, 2025 at 12:50 AM
‘Looks like a massive threat’: Analyst on Chinese #AI model @DeepSeek_AI https://buff.ly/4hRFa1l @ChannelNewsAsia @rwang0 of @constellationr tells CNA’s Asia First how big a threat DeepSeek poses to America’s tech industry.
Ollamaで14BモデルならDeepseek R1を動かせそうだし試してみたい💭
x.com/deepseek_ai/...
x.com/deepseek_ai/...
x.com
x.com
January 22, 2025 at 12:31 AM
Ollamaで14BモデルならDeepseek R1を動かせそうだし試してみたい💭
x.com/deepseek_ai/...
x.com/deepseek_ai/...
x.com
x.com
February 14, 2025 at 10:48 AM
This tweet appeared under this Techmeme headline:
Austin Carr / @austincarr:
Satya Nadella on @Microsoft's new "MAI-2" models, embracing @deepseek_ai, and competing with ChatGPT while supporting @OpenAI https://www.bloomberg.com/...
Austin Carr / @austincarr:
Satya Nadella on @Microsoft's new "MAI-2" models, embracing @deepseek_ai, and competing with ChatGPT while supporting @OpenAI https://www.bloomberg.com/...
May 17, 2025 at 6:38 AM
This tweet appeared under this Techmeme headline:
Austin Carr / @austincarr:
Satya Nadella on @Microsoft's new "MAI-2" models, embracing @deepseek_ai, and competing with ChatGPT while supporting @OpenAI https://www.bloomberg.com/...
Austin Carr / @austincarr:
Satya Nadella on @Microsoft's new "MAI-2" models, embracing @deepseek_ai, and competing with ChatGPT while supporting @OpenAI https://www.bloomberg.com/...