Image created with Ideogram 3.0. Image prompt: Lower-East-Side street-corner photograph reminiscent of a late-80s album cover: weathered red-brick tenement with exterior fire-escapes, canvas awning shading racks of vintage clothes; above the awning, a hand-painted board reads ‘OpenSource SPORTSWEAR’; a hanging blade sign in cursive script reads ‘OpenSource Boutique’; a crate of free floppy disks marked ‘Open-Source Software’ invites passersby; warm golden-hour light, subtle 35mm film grain, muted yet punchy color palette, gritty NYC vibe.
Jensen just announced NVIDIA’s Isaac GR00T N1.5 and GR00T-Dreams blueprint at COMPUTEX 2025: ⦿ Isaac GR00T N1.5 is the first update to NVIDIA’s open, generalized, fully customizable foundation model for humanoid reasoning and skills. ⦿ “Human demonstrations aren’t scalable — https://x.com/TheHumanoidHub/status/1924332201862414495
JUST IN🚨: Nvidia open sourced Physical AI models reasoning models that understand physical common sense and generate appropriate embodied decisions 👀 https://x.com/reach_vb/status/1924525937443365193
NVIDIA released new vision reasoning model for robotics: Cosmos-Reason1-7B 🤖 > first reasoning model for robotics 😱 > based on Qwen 2.5-VL-7B, use with @huggingface transformers or vLLM 🤗 > comes with SFT & alignment dataset and a new benchmark 👏 https://x.com/mervenoyann/status/1924817927561183498
We’re partnering with @Dell to accelerate secure, agentic enterprise AI solutions. Dell will be the first provider to offer our secure agents platform, Cohere North, to enterprises on-premises, which is crucial for regulated industries handling sensitive data 🧵 https://x.com/cohere/status/1924512634373865950
We’re partnering with @SAP to bring enterprise-ready agentic AI to businesses worldwide! Our models will be embedded into SAP Business Suite, offering secure and scalable AI capabilities. With Cohere’s cutting-edge models also available on SAP AI Core, enterprises can leverage https://x.com/cohere/status/1924858543716630644
QoL Update: Starting today, you will see an AI generated summary for all papers of Hugging Face Papers! 🔥 GG @mishig25 🐐 https://x.com/reach_vb/status/1925517801197879737
Alibaba’s Qwen team made Deep Research for Qwen Chat available for all users It’s pretty much like ChatGPT’s Deep Research, providing users the ability to prepare detailed reports on different subjects in a matter of minutes. https://x.com/adcock_brett/status/1924133804630753660
Great column by @htaneja & @FareedZakaria in @theinformation: « America’s historical technological leadership wasn’t built on protectionism and closed systems—it was fueled by creating a dynamic marketplace of optionality, including open platforms the world could build upon. Yet https://x.com/ClementDelangue/status/1924578324392587385
Meta just released KernelLLM 8B on Hugging Face ⚡ > On KernelBench-Triton Level 1, our 8B parameter model exceeds models such as GPT-4o and DeepSeek V3 in single-shot performance 🤯 > With multiple inferences, KernelLLM’s performance outperforms DeepSeek R1 https://x.com/reach_vb/status/1924478755898085552
Zed just dropped the fastest Agentic code editor built in Rust. Works with Claude Sonnet 3.7, Gemini 2.5 Pro and local models via Ollama. 100% opensource. https://x.com/Saboo_Shubham_/status/1921754009221906848
🚀 LangGraph Platform Now Supports MCP! Every deployed agent on LangGraph Platform now exposes its own MCP endpoint. Leverage your agents as tools in any client supporting streamable HTTP for MCP— no custom code or infrastructure required. 📚Docs: https://x.com/LangChainAI/status/1924863441862562279
Guide on how to Bridge @LangChainAI & @CamelAIOrg agents via the Model Context Protocol for seamless, cross-framework AI Agent collaboration. https://x.com/CamelAIOrg/status/1919750181622579627
Check out @shresbm and my blog on how get started building AI agents with Google Gemini and these awesome open-source tools: https://x.com/_philschmid/status/1924886346444710135
code agents > tool calling https://x.com/fdaudens/status/1923397074495627531
Let’s goo! Starting today you can access 5000+ LLMs powered by MLX directly from Hugging Face Hub! 🔥 All you need to do is click `Use this model` from any compatible model \o/ That’s it, all you need to get blazingly fast intelligence right at your terminal! What would you https://x.com/reach_vb/status/1924517049474101412
Wow, @jandotai is now Apache licensed – big win for on device community! 🔥 Way to go team! https://x.com/reach_vb/status/1925475572219568269
3. Agent factory: Foundry is the complete app platform for building apps and agents. We are adding support for more models from Grok, Hugging Face, Meta, Mistral, and more. Plus: Agentic retrieval in Azure AI Search, Foundry Agent Service, integration with Copilot Studio, and https://x.com/satyanadella/status/1924535900463366247
Devstral | Mistral AI https://mistral.ai/news/devstral
Meet Devstral, our SOTA open model designed specifically for coding agents and developed with @allhands_ai https://x.com/MistralAI/status/1925191937792901298
🚀 Build no-code agents with Open Agent Platform (OAP), our open-source, citizen developer platform for building, prototyping, and deploying agents. With Open Agent Platform, you can: 🔧 Build agents via a web UI— no heavy coding required 🧠 Connect to RAG servers for better https://x.com/LangChainAI/status/1925224206473842691
best agent chat ui, all open source!”” / X https://x.com/hwchase17/status/1924892270085448072
That’s a wrap on Interrupt 2025! 🚀 🌎 800 agent engineers from across the globe gathered in San Francisco for LangChain’s first industry conference to hear stories of teams building agents – and we’re still riding the high! @Cisco, @Uber, @Replit, @LinkedIn, @BlackRock,”” / X https://x.com/LangChainAI/status/1923089610772807959
This Thursday, the LlamaIndex team is hosting our first Discord office hours session! Drop in to ask anything LlamaIndex, and for an events driven agent workflows run-through and live coding session. See you there Thursday! Join the LlamaIndex Discord and add yourself to the https://x.com/llama_index/status/1924527932258845178
What has just been open-sourced by @Microsoft: ▪️ GitHub Copilot in Visual Studio Code ▪️ Natural Language Web (NL Web) ▪️ TypeAgent ▪️ Windows Subsystem for Linux (WSL) ▪️ Edit command-line text editor + Microsoft showed strong commitment to MCP as its standard open protocol https://x.com/TheTuringPost/status/1924598434507743728
🚀 Qwen Web Dev just got even better! ✨ One prompt. One website. One click to deploy. 💡 Let your creativity shine — and share it with the world. 🔥 What will you build today? https://x.com/Alibaba_Qwen/status/1924299942614688111
y’all know that @huggingface Spaces is the app store of AI what you don’t know is all these apps are MCP Servers thanks to @Gradio MCP server 😮 plug it to your favorite provider 🤠 insanely powerful! https://x.com/mervenoyann/status/1923406695000093095
MCP meets Ollama! In this video you’ll learn how build a 100% local MCP client that you can connect to any MCP server. 100% open-source code, step-by-step guide: https://x.com/akshay_pachaar/status/1921877497475485778
Hugging Face just dropped Tiny Agents into its own NPM package a squad of lightweight composable agents built on Hugging Face’s Inference Client and MCP stack https://x.com/_akhaliq/status/1924871432816783681
Really crazy! (open source 3D VR objects) https://x.com/fdaudens/status/1923095695843643848
Really cool how DeepSeek is now the benchmark for Nvidia”” / X https://x.com/teortaxesTex/status/1924588309688267139
📰 News in Arena: Mistral Medium 3 makes a strong debut with the community! Highlights: 💠 #11 overall in chat: a +90 point leap from Mistral Large 💠Top-tier in technical domains (#5 in Math, #7 in Hard Prompts & Coding) 💠#9 in WebDev Arena Congrats to @MistralAI on the https://x.com/lmarena_ai/status/1924482515244622120
Salesforce introduces: BLIP3-o: A Family of Fully Open Unified Multimodal Models—Architecture, Training and Dataset “”we introduce a novel approach that employs a diffusion transformer to generate semantically rich CLIP image features, in contrast to conventional VAE-based https://x.com/iScienceLuvr/status/1922843713514193076
Together AI and Agentica launched DeepCoder-14B-Preview, a code generation model that competes with top reasoning models like OpenAI’s o1 and DeepSeek-R1, but at a fraction of the size. Built on a 14 billion parameter Qwen model, DeepCoder uses a highly optimized reinforcement https://x.com/DeepLearningAI/status/1924570759793369303
🚀Applications are open for the Llama Startup Program!🚀 We’re thrilled to announce the Llama Startup Program, a new initiative designed to empower early-stage startups to innovate and build generative AI applications with Llama. Why Join the Llama Startup Program? ☑️Cloud https://x.com/AIatMeta/status/1925234408187175339
Enterprise Document AI & OCR | Mistral AI https://mistral.ai/solutions/document-ai
Salesforce just dropped BLIP3-o on Hugging Face A Family of Fully Open Unified Multimodal Models-Architecture, Training and Dataset https://x.com/_akhaliq/status/1923001183804764391
Hard to overstate the importance of Transformers (and datasets, tokenizers, etc) to the open-source and overall AI ecosystem. Can’t count the number of times I’ve personally used Transformers as a source-of-truth. Looking forward to more and deeper integrations with MLX + https://x.com/awnihannun/status/1923065749234647214
Great work @SullyOmarr and @HomamMalk Huge fans of @ottogrid_ai and am pumped to see it be a part of North from @cohere This is a 10/10 acquisition https://x.com/AtomSilverman/status/1923432542197211263
[2505.09343] Insights into DeepSeek-V3: Scaling Challenges and Reflections on Hardware for AI Architectures https://arxiv.org/abs/2505.09343
DeepSeek presents: Insights into DeepSeek-V3: Scaling Challenges and Reflections on Hardware for AI Architectures Elaborates on hardware architecture and model design in achieving cost-efficient large-scale training and inference https://x.com/arankomatsuzaki/status/1922844556430581761
Insights into DeepSeek-V3 Scaling Challenges and Reflections on Hardware for AI Architectures https://x.com/_akhaliq/status/1923001697498006016
Insights into DeepSeek-V3: Scaling Challenges and Reflections on Hardware for AI Architectures Overview: DeepSeek-V3 which is an LLM trained on 2,048 H800 GPUs, utilizes hardware-aware co-design incorporating Multi-head Latent Attention, MoE, FP8 training, and a Multi-Plane https://x.com/TheAITimeline/status/1924232113101890003
Designing models and hardware together — is it a new shift for the best cost-efficient models? This idea is used in DeepSeek-V3 that is trained on just 2,048 powerful NVIDIA H800 GPUs. A new research from @deepseek_ai clarifies how DeepSeek-V3 works using its key innovations: https://x.com/TheTuringPost/status/1924631209050833205
Do LLMs Really Understand Cell Biology? Interesting paper evaluating LLMs potential in understanding cell biology. Finding: It finds that specialist models don’t work so great. Generalist models, such as Qwen and DeepSeek, exhibit preliminary understanding capabilities within https://x.com/omarsar0/status/1922662317986099522
Everything you need to know to understand GRPO: GRPO (Group Relative Policy Optimization) is a reinforcement learning algorithm created by DeepSeek specifically for LLMs. It drops the need to use critic network like in PPO and so it doesn’t use absolute value estimate to https://x.com/TheTuringPost/status/1925146257372381485
This is a good overview of AI power use (small at individual level, big in aggregate). One thing that struck me: they tested LLama 3.1 405B and it averaged 3,353 joules per prompt. That is the equivalent of 2 minutes 50 seconds of human brain activity. https://x.com/emollick/status/1925178731389128744
Choosing Secure AI https://cohere.com/blog/choosing-secure-ai
Learn why enterprises are turning toward secure and private AI: https://x.com/cohere/status/1923083886319243518
google/shieldgemma-2-4b-it · Hugging Face https://huggingface.co/google/shieldgemma-2-4b-it
The Hugging Face Hub now auto-magically formats chat/reasoning messages in an interactive viewer 🪄 You can even toggle the reasoning blocks on/off, which is great for skipping those long-winded R1 traces 🥱 https://x.com/_lewtun/status/1924492654282207368
Microsoft and Hugging Face expand collaboration https://huggingface.co/blog/azure-ai-foundry
Multimodal model support is here in 0.7! Ollama now supports multimodal models via its new engine. Cool vision models to try👇 – Llama 4 Scout & Maverick – Gemma 3 – Qwen 2.5 VL – Mistral Small 3.1 and more 😍 Blog post 🧵👇 https://x.com/ollama/status/1923139667563528347
Two awesome new MLX + Hugging Face hub integrations. It’s easier than ever to get started running models locally: https://x.com/awnihannun/status/1924512714287939816
Analog Foundation Models “”In this work, we introduce a general and scalable method to robustly adapt LLMs for execution on noisy, low-precision analog hardware. Our approach enables state-of-the-art models – including Phi-3-mini-4k-instruct and Llama-3.2-1B-Instruct to – https://x.com/iScienceLuvr/status/1923269433751158884
ollama run devstral Devstral from @MistralAI and @allhands_ai is available on Ollama!”” / X https://x.com/ollama/status/1925198849263747147
Meet Document AI, our end-to-end document processing solution powered by the world’s best OCR model! https://x.com/MistralAI/status/1925577532595696116
BLIP3-o: A Family of Fully Open Unified Multimodal Models-Architecture, Training and Dataset Author’s Explanation: https://x.com/TheAITimeline/status/1924232118755824119
🚨 @frimelle and I are looking for a junior collaborator to research the Open Model Ecosystem! 🤖 Ideally, someone w/ AI/ML background, who can help w/ annotation pipeline + analysis. https://x.com/ShayneRedford/status/1925956405896307105
Andrew Ng is taking the stage at @Snowflake’s Dev Day 2025! Join us and fellow data professionals and AI builders to hear insights from a leading voice in AI. Plus, explore the Builders Hub, GenAI demos, and earn professional credentials. Save the date! 📅 June 5 | 📍 San https://x.com/DeepLearningAI/status/1924484108974993540
The fight for open models is a fight for freedom. For decades the free and open source movement has been fighting corporations and governments to protect user privacy, individual freedom, and public access to technology. Make no mistake, this is the same fight.”” / X https://x.com/BlancheMinerva/status/1925690741696651464
What would truly open-source AI look like? Not just open weights, open code/data, but *open development*, where the entire research and development process is public *and* anyone can contribute. We built Marin, an open lab, to fulfill this vision: https://x.com/percyliang/status/1924527490351169964
AM-Thinking-v1 looks like a strong 32B reasoning model. It outperforms DeepSeek-R1 and rivals Qwen3-235B-A22B. All built on top of open-source. The 32B scale is a great size for deployment and fine-tuning. Best part: the model is open-sourced! https://x.com/omarsar0/status/1922668488826741061
Qwen introduces: Parallel Scaling Law for Language Models “”We introduce the third and more inference-efficient scaling paradigm: increasing the model’s parallel computation during both training and inference time.”” “”We draw inspiration from classifier-free guidance (CFG)”” “”In https://x.com/iScienceLuvr/status/1923262107845525660
Qwen3 is abliterated! ✂️✂️✂️ What started as a weekend hack turned into three, but I’m happy with the result. Qwen3 was challenging with much stronger alignment and a new thinking mode that interfered with refusals. Here’s what I did to abliterate it https://x.com/maximelabonne/status/1924412611430404492
Lumina-Next on Qwen base, from Salesforce. Slightly surpasses Janus-Pro. I hope we start seeing actually multimodally pretrained unified models soon. https://x.com/teortaxesTex/status/1922961229233946869
You can now run Qwen3-32B on @HuggingFace with Cerebras Inference — and it’s ⚡️! Typing the question took longer than getting the answer 😅 https://x.com/fdaudens/status/1923107187284394368
Jensen: The humanoid robot is likely the only robot that will work – because technology needs scale, and most robots we’ve had so far are too low volume to drive the flywheel of technology improvements. The humanoid robot is likely to be the next multi-trillion-dollar industry. https://x.com/TheHumanoidHub/status/1924341417662672972
Larger models benefit less from strategic prompting. While strategies improve smaller models on long-text understanding and planning. Llama3.3-70B shows only marginal gains and, in some cases, experiences performance drops due to overcautious or inefficient reasoning paths.”” / X https://x.com/omarsar0/status/1924182839081218092
Qwen3 Technical Report Author’s Explanation: https://x.com/TheAITimeline/status/1924232110383960163
We’re starting to see more and more AI for chemistry and biology which I’m super excited about given the potential for good! @AIatMeta just released OMol25 on @huggingface, a dataset of 𝟭𝟬𝟬𝗠+ 𝗺𝗼𝗹𝗲𝗰𝘂𝗹𝗮𝗿 𝗰𝗼𝗻𝗳𝗼𝗿𝗺𝗲𝗿𝘀 spanning 83 elements and diverse chemical https://x.com/ClementDelangue/status/1924836697373565191
It was the week of video generation at @huggingface, on top of many new LLMs, VLMs and more! Let’s have a wrap 🌯 LLMs 💬 > Alibaba Qwen released WorldPM-72B, new World Preference Model trained with 15M preference samples (OS) > II-Medical-8B, new LLM for medical reasoning that https://x.com/mervenoyann/status/1924430139242283172
✨ All in One, Wan for All✨ We are excited to introduce our latest model to our talented community creators: Wan2.1-VACE, All-in-One Video Creation and Editing model. Model size: 1.3B, 14B License: Apache-2.0 📌 Wan2.1-VACE provides solutions for various tasks, including https://x.com/Alibaba_Wan/status/1922655324919779604
Bilibili dropped AniSORA on Hugging Face – anime video generation model capable of making manga, tuber, mad-style parodies and more! – Apache 2.0 licensed! 🔥 https://x.com/reach_vb/status/1924425789774123316
