DeepSeek releases open-weights math model with IMO gold medal performance https:// huggingface.co/deepseek-ai/Dee pSeek-Math-V2 # HackerNews # DeepSeek # OpenWeights # Math # Model # IMO # Gold # Medal # Performance # AI # Innovation # MachineLearning
Interest | Match | Feed
Interest | Match | Feed
Origin
mastodon.social
December 1, 2025 at 10:21 AM
Try AllenAI's OLMo 3-Think (32B) for free, no login needed
olmo.gptchatly.com
#olmo #allen #allenai #ai #OpenSource #openweights #LLMs
olmo.gptchatly.com
#olmo #allen #allenai #ai #OpenSource #openweights #LLMs
November 24, 2025 at 7:43 AM
Try AllenAI's OLMo 3-Think (32B) for free, no login needed
olmo.gptchatly.com
#olmo #allen #allenai #ai #OpenSource #openweights #LLMs
olmo.gptchatly.com
#olmo #allen #allenai #ai #OpenSource #openweights #LLMs
Self-Hosting LLMs: Why the Details Matter 🔍
More teams turn to self-hosted LLMs for data protection, lower API costs, and digital sovereignty. One key point often missed: #OpenSource vs. #OpenWeights. What companies should know: 🔗 https://zammad.com/en/blog/self-hosting-llms
#LLM #SelfHosting
More teams turn to self-hosted LLMs for data protection, lower API costs, and digital sovereignty. One key point often missed: #OpenSource vs. #OpenWeights. What companies should know: 🔗 https://zammad.com/en/blog/self-hosting-llms
#LLM #SelfHosting
November 20, 2025 at 11:31 AM
Self-Hosting LLMs: Why the Details Matter 🔍
More teams turn to self-hosted LLMs for data protection, lower API costs, and digital sovereignty. One key point often missed: #OpenSource vs. #OpenWeights. What companies should know: 🔗 https://zammad.com/en/blog/self-hosting-llms
#LLM #SelfHosting
More teams turn to self-hosted LLMs for data protection, lower API costs, and digital sovereignty. One key point often missed: #OpenSource vs. #OpenWeights. What companies should know: 🔗 https://zammad.com/en/blog/self-hosting-llms
#LLM #SelfHosting
Kimi K2 Thinking is impressive, it is as at least as good as other top foundation models, fully openweights, 1 trillion parameters but only using 32B in a MoE setting. Also you can download and reuse it for free.
github.com/MoonshotAI/K...
#ai #kimik2thinking
github.com/MoonshotAI/K...
#ai #kimik2thinking
GitHub - MoonshotAI/Kimi-K2: Kimi K2 is the large language model series developed by Moonshot AI team
Kimi K2 is the large language model series developed by Moonshot AI team - MoonshotAI/Kimi-K2
github.com
November 8, 2025 at 1:12 PM
Kimi K2 Thinking is impressive, it is as at least as good as other top foundation models, fully openweights, 1 trillion parameters but only using 32B in a MoE setting. Also you can download and reuse it for free.
github.com/MoonshotAI/K...
#ai #kimik2thinking
github.com/MoonshotAI/K...
#ai #kimik2thinking
I used the new ChatGPT openweights model to help me write up a Gemini Gem profile to describe Sailor Moon outfits. I left the instructions open so I should be able to change characters and outfit pieces more easily. I had to test with Sunset Shimmer with no shoes because I have a weakness.
August 8, 2025 at 1:11 AM
I used the new ChatGPT openweights model to help me write up a Gemini Gem profile to describe Sailor Moon outfits. I left the instructions open so I should be able to change characters and outfit pieces more easily. I had to test with Sunset Shimmer with no shoes because I have a weakness.
The world doesn't need another open-weights model. Especially not one that seems to be behind its peers.
#openweights #AI
www.theguardian.com/...
#openweights #AI
www.theguardian.com/...
OpenAI takes on Meta and DeepSeek with free and customisable AI models
Developer of ChatGPT says new tools will be ‘for wide benefit’, echoing announcement by Mark Zuckerberg
www.theguardian.com
August 6, 2025 at 6:00 PM
The world doesn't need another open-weights model. Especially not one that seems to be behind its peers.
#openweights #AI
www.theguardian.com/...
#openweights #AI
www.theguardian.com/...
The 120B model reaches reasoning performance on par with o4-mini—while running on a single 80 GB GPU.
The compact 20B model fits on laptops with ~16 GB RAM.
Both support chain-of-thought, tool use, and agentic workflows—fully local and commercially usable.
#OpenWeights #LLM #Apache2
The compact 20B model fits on laptops with ~16 GB RAM.
Both support chain-of-thought, tool use, and agentic workflows—fully local and commercially usable.
#OpenWeights #LLM #Apache2
OpenAI takes on Meta and DeepSeek with free and customisable AI models
Developer of ChatGPT says new tools will be ‘for wide benefit’, echoing announcement by Mark Zuckerberg
www.theguardian.com
August 6, 2025 at 7:54 AM
The 120B model reaches reasoning performance on par with o4-mini—while running on a single 80 GB GPU.
The compact 20B model fits on laptops with ~16 GB RAM.
Both support chain-of-thought, tool use, and agentic workflows—fully local and commercially usable.
#OpenWeights #LLM #Apache2
The compact 20B model fits on laptops with ~16 GB RAM.
Both support chain-of-thought, tool use, and agentic workflows—fully local and commercially usable.
#OpenWeights #LLM #Apache2
gpt-oss vient de sortir, le modèle openweights de #Openai il est dispo sur #azure #ai foundry et supporté par Local Foundry
azure.microsoft.com/en-us/blog/o... sous licence Apache 2.0. Tous les détails technique sur huggingface.co/blog/welcome...
azure.microsoft.com/en-us/blog/o... sous licence Apache 2.0. Tous les détails technique sur huggingface.co/blog/welcome...
OpenAI’s open‑source model: gpt‑oss on Azure AI Foundry and Windows AI Foundry | Microsoft Azure Blog
OpenAI’s gpt‑oss models gives developers and enterprises the ability to run, adapt, and deploy OpenAI models on their own terms. Learn more.
azure.microsoft.com
August 6, 2025 at 7:42 AM
gpt-oss vient de sortir, le modèle openweights de #Openai il est dispo sur #azure #ai foundry et supporté par Local Foundry
azure.microsoft.com/en-us/blog/o... sous licence Apache 2.0. Tous les détails technique sur huggingface.co/blog/welcome...
azure.microsoft.com/en-us/blog/o... sous licence Apache 2.0. Tous les détails technique sur huggingface.co/blog/welcome...
OpenAI released its first open-weight models since GPT-2:
gpt-oss-120b and gpt-oss-20b can run locally and be fine-tuned.
A notable shift—away from fully proprietary releases.
Why now? Competition, regulation, community pressure—or strategic repositioning?
#OpenWeights #OpenAI #AI #LLM #AIstrategy
gpt-oss-120b and gpt-oss-20b can run locally and be fine-tuned.
A notable shift—away from fully proprietary releases.
Why now? Competition, regulation, community pressure—or strategic repositioning?
#OpenWeights #OpenAI #AI #LLM #AIstrategy
Introducing gpt-oss
gpt-oss-120b and gpt-oss-20b push the frontier of open-weight reasoning models
openai.com
August 6, 2025 at 7:15 AM
OpenAI released its first open-weight models since GPT-2:
gpt-oss-120b and gpt-oss-20b can run locally and be fine-tuned.
A notable shift—away from fully proprietary releases.
Why now? Competition, regulation, community pressure—or strategic repositioning?
#OpenWeights #OpenAI #AI #LLM #AIstrategy
gpt-oss-120b and gpt-oss-20b can run locally and be fine-tuned.
A notable shift—away from fully proprietary releases.
Why now? Competition, regulation, community pressure—or strategic repositioning?
#OpenWeights #OpenAI #AI #LLM #AIstrategy
A key distinction: "open-weight" isn't "open-source." These OpenAI models lack training data and have usage restrictions, preventing true community contribution and transparency. This impacts the open AI ecosystem. #OpenWeights 4/6
August 6, 2025 at 7:00 AM
A key distinction: "open-weight" isn't "open-source." These OpenAI models lack training data and have usage restrictions, preventing true community contribution and transparency. This impacts the open AI ecosystem. #OpenWeights 4/6
🚀 OpenAI drops #gptoss: open-weight language models.
openai.com/index/introd...
#AI #opensource #ML #tech #OpenWeights
openai.com/index/introd...
#AI #opensource #ML #tech #OpenWeights
Introducing gpt-oss
gpt-oss-120b and gpt-oss-20b push the frontier of open-weight reasoning models
openai.com
August 5, 2025 at 7:32 PM
🚀 OpenAI drops #gptoss: open-weight language models.
openai.com/index/introd...
#AI #opensource #ML #tech #OpenWeights
openai.com/index/introd...
#AI #opensource #ML #tech #OpenWeights
No fortune cookie needed: Your Tokens May Vary www.exponentialview.co/p/chinas-dif... #Scale #Application #Integration #Pragmatic #Efficiency #OpenWeights
🔮 China’s on a different AI path
America bets on AGI. China bets on margins.
www.exponentialview.co
July 27, 2025 at 7:03 AM
No fortune cookie needed: Your Tokens May Vary www.exponentialview.co/p/chinas-dif... #Scale #Application #Integration #Pragmatic #Efficiency #OpenWeights
Finally, the press is reporting on the issue with Fake Open-Source AI, which developers have warned against for some time. Words have meaning, and you cannot simply redefine them.
#AI #opensource #openweights
dig.watch/updates/me...
#AI #opensource #openweights
dig.watch/updates/me...
Meta faces backlash over open source AI claims | Digital Watch Observatory
OpenUK’s Amanda Brock says Meta’s licensing breaks key principles of open source and creates unnecessary friction for developers.
dig.watch
May 30, 2025 at 9:53 PM
Finally, the press is reporting on the issue with Fake Open-Source AI, which developers have warned against for some time. Words have meaning, and you cannot simply redefine them.
#AI #opensource #openweights
dig.watch/updates/me...
#AI #opensource #openweights
dig.watch/updates/me...
#OpenAI signals it plans to release a new open-weight AI model - joining #Meta and #Mistral in the race to make #AI more open and accessible. Great rundown by @lizziegibney.bsky.social . #OpenWeights #LLM #AGI #ResearchSky ‘AI models are capable of n...
| Details | Interest | Feed |
| Details | Interest | Feed |
Origin
bsky.brid.gy
May 16, 2025 at 5:16 PM
#OpenAI signals it plans to release a new open-weight AI model - joining #Meta and #Mistral in the race to make #AI more open and accessible. Great rundown by @lizziegibney.bsky.social.
#OpenWeights #LLM #AGI #ResearchSky
#OpenWeights #LLM #AGI #ResearchSky
‘AI models are capable of novel research’: OpenAI’s chief scientist on what to expect
Jakub Pachocki, who leads the firm’s development of advanced models, is excited to release an open version to researchers.
www.nature.com
May 16, 2025 at 5:12 PM
#OpenAI signals it plans to release a new open-weight AI model - joining #Meta and #Mistral in the race to make #AI more open and accessible. Great rundown by @lizziegibney.bsky.social.
#OpenWeights #LLM #AGI #ResearchSky
#OpenWeights #LLM #AGI #ResearchSky
🧠 Researchers: run it
🧰 Engineers: fine-tune it
🧪 Builders: break it
Tell us what you find.
Apriel-5B models are permissively licensed (MIT) and ready to chat.
#Apriel #LLM #AI #OpenWeights #FastLLM #SLAM #ServiceNow #ServiceNowResearch
🧰 Engineers: fine-tune it
🧪 Builders: break it
Tell us what you find.
Apriel-5B models are permissively licensed (MIT) and ready to chat.
#Apriel #LLM #AI #OpenWeights #FastLLM #SLAM #ServiceNow #ServiceNowResearch
April 11, 2025 at 8:15 PM
🧠 Researchers: run it
🧰 Engineers: fine-tune it
🧪 Builders: break it
Tell us what you find.
Apriel-5B models are permissively licensed (MIT) and ready to chat.
#Apriel #LLM #AI #OpenWeights #FastLLM #SLAM #ServiceNow #ServiceNowResearch
🧰 Engineers: fine-tune it
🧪 Builders: break it
Tell us what you find.
Apriel-5B models are permissively licensed (MIT) and ready to chat.
#Apriel #LLM #AI #OpenWeights #FastLLM #SLAM #ServiceNow #ServiceNowResearch
Meta promised a 10M-token AI. The reality? Junk summaries, hardware nightmares, and disappointed devs. Here's what went wrong with Llama 4. #Llama4 #AIhype #OpenWeights
Meta’s Llama 4 Hype Hits a Wall of Reality
Meta promised a 10M-token AI revolution with Llama 4. Instead, users got tech headaches and limited performance. Is the hype breaking the model?
geekoo.news
April 10, 2025 at 9:01 AM
Meta promised a 10M-token AI. The reality? Junk summaries, hardware nightmares, and disappointed devs. Here's what went wrong with Llama 4. #Llama4 #AIhype #OpenWeights
There haven't been many times where I have said something positive about OpenAI, but they managed to call their upcoming AI model open-weights instead of pretending it's open-source like so many others.
#opensource #openweights #AI
siliconangle.com/202...
#opensource #openweights #AI
siliconangle.com/202...
OpenAI to launch its first 'open-weights' model since 2019 - SiliconANGLE
OpenAI to launch its first 'open-weights' model since 2019 - SiliconANGLE
siliconangle.com
April 2, 2025 at 2:38 AM
There haven't been many times where I have said something positive about OpenAI, but they managed to call their upcoming AI model open-weights instead of pretending it's open-source like so many others.
#opensource #openweights #AI
siliconangle.com/202...
#opensource #openweights #AI
siliconangle.com/202...
OpenAI plans to release an open-weight AI model this year, challenging DeepSeek R1 and Meta Llama. Run it on your own hardware, customize for sensitive tasks, and join early access. The open AI race heats up. #OpenAI #AIModels #OpenWeights
April 1, 2025 at 3:01 AM
OpenAI plans to release an open-weight AI model this year, challenging DeepSeek R1 and Meta Llama. Run it on your own hardware, customize for sensitive tasks, and join early access. The open AI race heats up. #OpenAI #AIModels #OpenWeights
I would say it wasn't openAI that stalled open-source AI developments. Instead, Meta and Co.'s OpenWeights models have distorted the open-source community and made it impossible to cut through the noise to find open-source models to contribute to.
#opensource #AI
www.ccn.com/news/tec...
#opensource #AI
www.ccn.com/news/tec...
Mistral CEO Says OpenAI Disrupted Open-Source Momentum, Cites Nvidia Partnership as Blueprint for Progress | CCN.com
Mistral CEO Arthur Mensch suggested that open-source AI development stalled after OpenAI released GPT-3 in 2020.
www.ccn.com
March 24, 2025 at 5:59 PM
I would say it wasn't openAI that stalled open-source AI developments. Instead, Meta and Co.'s OpenWeights models have distorted the open-source community and made it impossible to cut through the noise to find open-source models to contribute to.
#opensource #AI
www.ccn.com/news/tec...
#opensource #AI
www.ccn.com/news/tec...
The paranoia around openweights/source AI is high, last night I had to debate a guy who thought that .safetensors could indeed contain a "chinese malware"...
I had to keep explaning that safetensors are like CSV files... a collection of numbers unlikely to contain executable code.
I had to keep explaning that safetensors are like CSV files... a collection of numbers unlikely to contain executable code.
February 2, 2025 at 4:41 PM
The paranoia around openweights/source AI is high, last night I had to debate a guy who thought that .safetensors could indeed contain a "chinese malware"...
I had to keep explaning that safetensors are like CSV files... a collection of numbers unlikely to contain executable code.
I had to keep explaning that safetensors are like CSV files... a collection of numbers unlikely to contain executable code.
Neither Deepseek or Metas models are opensource. They are openweights. Don’t let them redefine what open source means.
Friends, for something to be open source, we need to see
1. The data it was trained & evaluated on
2. The code
3. Model architecture
4. Model weights.
DeepSeek only gives 3, 4. And I'll see the day that anyone gives us #1 without being forced to do so, because all of them are stealing data.
1. The data it was trained & evaluated on
2. The code
3. Model architecture
4. Model weights.
DeepSeek only gives 3, 4. And I'll see the day that anyone gives us #1 without being forced to do so, because all of them are stealing data.
January 29, 2025 at 8:34 PM
Neither Deepseek or Metas models are opensource. They are openweights. Don’t let them redefine what open source means.
Para ser más específicos (porque sí importa), #DeepSeek *no* es de código abierto: es de #OpenWeights con la diferencia de que puedes ejecutarlo en tu propio ordenador con el código, pero no sabes exactamente con qué material fue entrenado. Una distinción importante que vale la pena enfatizar'👇
Also, to be specific here (because it does matter) DeepSeek is *not* open source — it’s open _weights_ with the difference being that you can run it on your own machine with the code, but you don’t know what material exactly it was trained on. An important distinction, worth emphasizing.
January 27, 2025 at 4:26 PM
Para ser más específicos (porque sí importa), #DeepSeek *no* es de código abierto: es de #OpenWeights con la diferencia de que puedes ejecutarlo en tu propio ordenador con el código, pero no sabes exactamente con qué material fue entrenado. Una distinción importante que vale la pena enfatizar'👇
Our CEO @amandabrock.bsky.social comments to @forbes.com on the launch of an Open Weights Definition by the Open Source Alliance https://buff.ly/40wMdFy in the run-up to AI Openness Day at SOOCon25 on 5th Feb and the French-led AI Action Summit on 10 Feb #AI #openweights #openmodels #opensource
January 24, 2025 at 12:00 PM
Our CEO @amandabrock.bsky.social comments to @forbes.com on the launch of an Open Weights Definition by the Open Source Alliance https://buff.ly/40wMdFy in the run-up to AI Openness Day at SOOCon25 on 5th Feb and the French-led AI Action Summit on 10 Feb #AI #openweights #openmodels #opensource