Image created with OpenAI gpt-image-1. Image prompt: Single-panel cartoon with loose, hand‑inked lines, bean‑bodied figures, muted flat colors, minimal props, and deadpan humor: Barn‑raising scene. Farm animals hammer code‑filled planks onto a communal barn while a chicken offers a pull request. Large bold title text centered at top: “OPENSOURCE” Muted colors, flat shading, black ink outlines. 16:9. Caption: “Warranty limited to eggshell cracks.”
“Nvidia just released Describe Anything 3B – Multimodal LLM for Detailed Localized Image and Video Captioning ⚡ > integrates full-image/ video context with fine-grained local details using a focal prompt and a localised vision backbone with gated cross-attention DAM-3B > https://x.com/reach_vb/status/1914962078571356656
“Nvidia presents Eagle 2.5! – A family of frontier VLMs for long-context multimodal learning – Eagle 2.5-8B matches the results of GPT-4o and Qwen2.5-VL-72B on long-video understanding https://x.com/arankomatsuzaki/status/1914517474370052425
“Wait, Nvidia dropped a 4 MILLION context length Llama 3.1 Nemotron 🤯 could literally drop entire codebases in it! https://x.com/reach_vb/status/1912743420851875986
“We just dropped a new SoTA lipsync model on @FAL: Hummingbird-0 Available now as a research preview, it’s the most accurate zero-shot lipsync model we’ve tested, open or closed source. https://x.com/heytavus/status/1915435703833641231
“Adobe announced DRAGON on Hugging Face Distributional Rewards Optimize Diffusion Generative Models https://x.com/_akhaliq/status/1914602497148154226
“Spotify just announced ViSMaP on Hugging Face Unsupervised Hour-long Video Summarisation by Meta-Prompting https://x.com/_akhaliq/status/1915703054701044209
“LiveCC just dropped on Hugging Face Learning Video LLM with Streaming Speech Transcription at Scale video LLM capable of real-time commentary, trained with a novel video-ASR streaming method, SOTA on both streaming and offline benchmarks. https://x.com/_akhaliq/status/1915094398364197101
“New short course: Building Code Agents with Hugging Face smolagents! Learn how to build code agents in this course, created in collaboration with @huggingface, and taught by @Thom_Wolf, its co-founder and CSO, and @AymericRoucher, Hugging Face’s Project Lead on Agents. https://x.com/AndrewYNg/status/1915101920500564406
“If you can’t shell out 2K$ 😱 to learn about LLM evaluations, take a look at our free/open resources: 1. LLM guidebook: From theory to troubleshooting https://x.com/clefourrier/status/1915339216344526896
Distilling DeepSeek-R1 intelligence into local models with HP AI Studio
https://reinvent.hp.com/ai-studio-may-1
MiniLLM (MiniLLM) https://huggingface.co/MiniLLM
OmDet-Turbo https://huggingface.co/docs/transformers/main/en/model_doc/omdet-turbo
“NEW: You can now use Dia 1.6B SoTA Text-to-Speech model directly on Hugging Face via @FAL 🔥 You can get up-to 25 generations for less than a dollar 🤗 Run it 5 lines of code too: import requests API_URL = “https://router.huggingface. co/fal-ai/fal-ai/dia-tts” headers = { https://x.com/reach_vb/status/1915418386818834792
“We shipped an alpha version of the new Surya OCR model. No hype, just facts: – 90+ languages (focus on en, romance langs, zh, ar, ja, ko) – LaTeX and formatting – Char/word/line bboxes – ~500M non-embed params – 10-20 pages/s https://x.com/VikParuchuri/status/1915492483955384659
“Perception Encoder models and datasets: https://x.com/mervenoyann/status/1915723397272654194
sand-ai/MAGI-1 · Hugging Face https://huggingface.co/sand-ai/MAGI-1
“OpenRLHF is a pioneering framework to use vLLM for RLHF, driving many design and implementation of vLLM’s features for RLHF, making vLLM a popular choice for many RLHF frameworks. Learn more about the story at https://x.com/vllm_project/status/1915307134256091570
“Kortix AI dropped Suna, an open-source, general AI Agent Akin to an “AI employee,” it can reason, plan, and act across domains using a virtual computer Can do a lot, including writing files, executing code, browsing the web, and using the terminal! https://x.com/rowancheung/status/1914930137465843813
“TextArena went live on Hugging Face It’s an open-source collection of competitive text-based games for LLMs, spanning 57+ unique environments Tests for different agentic behaviors—negotiation, theory of mind, deception, via competitive play https://x.com/rowancheung/status/1914567435228795391
“Here is a new open-source IDE to help you build multi-agent systems. It’s like Cursor but specifically for building multi-agent workflows. It’s powered by OpenAI Agents SDK, connects MCP servers, and can integrate into your apps using HTTP or the SDK. https://x.com/omarsar0/status/1915040973601776092
“Introducing Kortix Suna: the world’s first Open-Source General AI Agent built to work like a Human. https://x.com/kortixai/status/1914727901573927381
“We’ve open-sourced our MCP (Model Context Protocol) — now available on GitHub. Plug it into your favorite client and start building richer AI integrations with Notion in minutes ✨ https://x.com/NotionAPI/status/1910043694520046078
“Fuck it, starting today you can run inference across 30,000+ Flux and SDXL LoRAs on the Hugging Face Hub via Inference Providers (powered by @FAL ⚡) And.. it gets better, you can generate over 40+ images in less than A DOLLAR! Go try it now on your favourite LoRA on HF 🤗 https://x.com/reach_vb/status/1915830938438717777
“Llama 4 Maverick illustrates a key challenge in calculating AI training compute: How to account for when a smaller model is trained using a larger model’s outputs? We’re updating our methodology and removing Maverick from our list of models exceeding 1e25 FLOP. Here’s why… 🧵 https://x.com/EpochAIResearch/status/1913329195171688742
“Nvidia just dropped Describe Anything on Hugging Face Detailed Localized Image and Video Captioning https://x.com/_akhaliq/status/1914917564137828622
BMW to integrate DeepSeek AI in its new vehicles in China later this year | Reuters https://www.reuters.com/business/autos-transportation/bmw-integrate-deepseek-ai-its-new-vehicles-china-later-this-year-2025-04-23/
“vLLM🤝🤗! You can now deploy any @huggingface language model with vLLM’s speed. This integration makes it possible for one consistent implementation of the model in HF for both training and inference. 🧵 https://x.com/vllm_project/status/1912958639633277218
“Special call out for help: We’ve been in contact with the @deepseek_ai team to potentially do a talk (in Chinese) but unfortunately the… situation… has changed We -still- want good DeepSeek infra+model speakers – ofc you don’t have to work at deepseek to do this! Please https://x.com/swyx/status/1913818917765870050
“«DeepSeek to remain open-source to benefit the world» – Zhang Hanhui, the Chinese Ambassador to Russia, for the Russian state-owned news agency TASS, 14th Apr 2025 I didn’t know ambassadors get to decide this, but thanks man! https://x.com/teortaxesTex/status/1914204333857509488
Chat UI Energy Score – a Hugging Face Space by jdelavande https://huggingface.co/spaces/jdelavande/chat-ui-energy
“1/3 🚀Thrilled to introduce Wan2.1-FLF2V-14B – our first 14B-parameter large model for First-Last-Frame to video generation! Open-source, open-source, open-source! Empowering digital artists with unprecedented efficiency and creative flexibility. #wan #AIGC #alart https://x.com/Alibaba_Wan/status/1912874582635397233
“@AtomSilverman @AgentOpsAI @grok remix as a showcase article. https://x.com/kidehen/status/1912976623818907721
“NYU researchers have introduced RUKA, an open‑source, tendon‑driven robotic hand with 15 DOF that costs only $1.3 k and can operate for 20 straight hours without any performance loss. It learns joint‑to-actuator and fingertip‑to‑actuator models from motion‑capture data. https://x.com/TheHumanoidHub/status/1913295130259525823
“Finetuning on raw DeepSeek R1 reasoning traces makes models overthink. One of our early s1 versions was overthinking so much, it questioned the purpose of math when just asking what’s 1+1😁 Retro-Search by @GXiming & team reduces overthinking + improves performance!” / X https://x.com/Muennighoff/status/1914768451618660782
“Perplexity serves MoEs like post-trained versions of DeepSeek-v3. These models can be made to utilize GPUs efficiently in multi-node settings, achieving high throughput and low latency simultaneously, compared to single-node deployments. https://x.com/AravSrinivas/status/1913309684397908399
Rivian elects Cohere’s CEO to its board in latest signal the EV maker is bullish on AI | TechCrunch https://techcrunch.com/2025/04/21/rivian-elects-coheres-ceo-to-its-board-in-latest-signal-the-ev-maker-is-bullish-on-ai/
“So cool to see transformers becoming the source of truth for model definition & collaborating with wonderful partners like vLLM to have these models run everywhere the fastest! As a model builder, it means that you integrate with Hugging Face & instantly get hundreds of https://x.com/ClementDelangue/status/1914432076956262495
“I can’t believe that DeepSeek predicted this… https://x.com/teortaxesTex/status/1915055072532402686
“Using diverse Natural Language Processing models requires navigating multiple complex frameworks and writing repetitive code. Langformers simplifies this by providing an open-source Python library. It offers a unified, factory-based interface for various LLM and Masked Language https://x.com/rohanpaul_ai/status/1915650732709269750
“SRPO: A Cross-Domain Implementation of Large-Scale Reinforcement Learning on LLM “In this work, we present two-Staged history-Resampling Policy Optimization (SRPO), which successfully surpasses the performance of DeepSeek-R1-Zero-32B on the AIME24 and LiveCodeBench benchmarks. https://x.com/iScienceLuvr/status/1914622980296192357
“open-source tool to expose your FastAPI endpoints as Model Context Protocol tools with zero config. Simple. Flexible. Production-ready. https://x.com/GithubProjects/status/1910347261000773933
“Alibaba released Wan 2.1-FLF2V-14B, an open-source AI for video generation What makes this model unique is its ability to let users control their generations with first and last frame inputs The result is 720p cinematic clips with smooth transitions! https://x.com/rowancheung/status/1914201312222118346
“We’re publishing new queryable datasets to help researchers explore interpretable features in DeepSeek R1. https://x.com/GoodfireAI/status/1915802798513598490
cohere on X: “Introducing Embed 4: our latest state-of-the-art multimodal embedding model that enables enterprises to securely add powerful search and retrieval capabilities to their agentic AI applications! https://t.co/eTpgA22P7k” / X
https://x.com/cohere/status/1912128813104078999
“ByteDance announces Vidi on Hugging Face Large Multimodal Models for Video Understanding and Editing https://x.com/_akhaliq/status/1914925322413264937
“Meta released WebSSL DINO & ViT models on Hugging Face 300M to 7B 🔥 Notes: > Visual SSL outperforms CLIP on Vision-Centric VQA and closes the gap on OCR & Chart tasks when scaled properly > CLIP saturates at 3B parameters, while SSL shows log-linear improvements up to 7B+ > https://x.com/reach_vb/status/1915453821251375552
“Fully sharded systems use fixed strategies, ignoring dynamic memory changes during training. DeepCompile compiles models into graphs, using profiling-guided passes to flexibly time operations based on runtime memory. It boosts Llama 3 70B/Mixtral 8x7B training up to https://x.com/rohanpaul_ai/status/1914866314122015149
“ByteDance just announced QuaDMix on Hugging Face Quality-Diversity Balanced Data Selection for Efficient LLM Pretraining https://x.com/_akhaliq/status/1915656590130036887
“The Qwen Chat APP is now available for both iOS and Android users! It’s free to use and designed to assist with creativity, collaboration, and endless possibilities. Just ask, and let Qwen Chat handle the rest. Scan the QR code to quickly access the Qwen Chat APP! https://x.com/Alibaba_Qwen/status/1915761990703697925
“LlamaIndex’s integration with @milvusio now supports full-text search with BM25! Full-text search allows hybrid search for RAG pipelines, combining the power of vector search and traditional keyword matching. Check out how to use the integration in this tutorial: https://x.com/llama_index/status/1914815391798534571
