Hugging Face
89 analyses mentioning Hugging Face, newest first. Page 2 of 2.
Hugging Face Introduces Mutable Storage Buckets for ML Artifacts
Hugging Face has launched Storage Buckets, a new S3-like object storage solution built on its Xet backend, designed for efficiently managing the constantly changing intermediate files generated during machine learning training.
Snowflake Unveils Ulysses: A New Approach to Long Sequence Training
Snowflake AI Research has introduced Ulysses, a sequence parallelism protocol designed to efficiently train large language models on sequences of millions of tokens. Utilizing all-to-all communication and partitioning attention heads across GPUs, Ulysses overcomes the memory limitations of traditional attention mechanisms, opening the door to more capable and complex models.
European Startup Multiverse Computing Optimizes LLMs with Compressed Models
Multiverse Computing, a Spanish startup, is releasing compressed versions of large language models, aiming to make AI deployment more accessible to companies.
Unsloth Makes Small LLM Fine-Tuning Accessible (But Hype is Overstated)
This blog post details how Unsloth and Hugging Face Jobs enable developers to efficiently fine-tune small language models like LiquidAI/LFM2.5-1.2B-Instruct, leveraging cloud GPU resources and coding agents for automation.
Synthetic Personas: A Data Wall Breaker for Japan’s AI
Japanese AI development is facing a critical data scarcity issue, but NTT DATA’s research demonstrates a novel solution: using synthetic data generated by Nemotron-Personas-Japan to dramatically boost model accuracy and performance while preserving privacy.
AI Agents Now Automate CUDA Kernel Development
A new GitHub skill allows coding agents, like Codex and Claude, to automatically generate optimized CUDA kernels for machine learning models, simplifying a traditionally complex development process.
OpenEnv: Testing Real-World AI Agents Through Complex Calendars
OpenEnv, a new framework from Meta and Hugging Face, provides a production-grade calendar management environment (the Calendar Gym) for evaluating AI agents under realistic constraints, revealing critical limitations in agent reliability and highlighting the need for robust multi-step reasoning.
Transformers.js v4 Preview Released: WebGPU Acceleration and Modular Updates
Hugging Face has released Transformers.js v4 (preview), boasting significant performance improvements through a new WebGPU runtime, a modular codebase, and expanded model support. This update dramatically accelerates transformer model execution, now fully compatible with browser and server-side JavaScript environments.
Mistral AI's Privacy-Focused Speech Models Challenge OpenAI
Paris-based startup Mistral AI has launched two new speech-to-text models, Voxtral Transcribe 2 and Voxtral Realtime, designed for faster, more accurate, and cheaper transcription while prioritizing on-device processing and data privacy, directly competing with OpenAI's offerings.
Hugging Face Decentralizes Model Evaluation with 'Community Evals'
Hugging Face is launching 'Community Evals,' a new initiative to decentralize model evaluation by allowing the entire community to report scores for benchmarks, aggregating results from various sources into a centralized hub.
H Company's Holo2-235B-A22B-Preview Model Shatters UI Localization Benchmarks
H Company unveils Holo2-235B-A22B-Preview, a new UI localization model achieving record accuracy thanks to agentic localization and leveraging SkyPilot for scalable training.
China's Open-Source AI Ecosystem Rapidly Takes Shape, Driven by 'DeepSeek Moments'
China’s open-source AI community is experiencing a significant shift, fueled by organizations like Alibaba, Tencent, and ByteDance embracing open-source models and infrastructure, creating a dynamic ecosystem that’s redefining global AI development.
TechCrunch Doubles Down: Founder Summit 2026 Promises Scaling Insights
TechCrunch is hosting its flagship founder conference, TechCrunch Founder Summit 2026, in Boston, offering a packed schedule of sessions focused on growth, execution, and real-world scaling strategies for startups.
Claude Powers Open Source Kernel Generation: A New Upskilling Approach
Researchers have successfully utilized Anthropic's Claude model to generate CUDA kernels for open-source models, demonstrating a new approach to upskilling coding agents on complex tasks.
China's Open-Source AI Ecosystem: A Hardware-First Shift
China's open-source AI community, spurred by the 'DeepSeek Moment,' is rapidly evolving beyond purely model-centric approaches, prioritizing hardware integration, sustainable deployment, and a shift towards smaller, more practical models – fundamentally altering its trajectory in the global AI landscape.
Foundation Models Revolutionize Time Series Forecasting in 2026
The 2026 Time Series Toolkit reveals five key foundation models—Chronos-2, MOIRAI-2, Lag-Llama, Time-LLM, and TimesFM—that are fundamentally changing how businesses approach time series forecasting.
AI Hallucinations Plague Prestigious NeurIPS Conference Papers
An AI detection startup, GPTZero, found 100 instances of fabricated citations across 51 papers submitted to the NeurIPS conference, raising concerns about the accuracy of AI-generated content within leading academic research.
China's Open AI Surge: A Turning Point for the Ecosystem
One year after the release of DeepSeek R1, a dramatic shift has occurred in China’s AI landscape, fueled by an explosion of open-source models and a strategic realignment towards leveraging open-source capabilities for competitive advantage.
Runpod's Wild Ride: From Basement Mining to $120M Revenue
Runpod, an AI app hosting platform, has experienced remarkable growth, hitting a $120 million annual revenue run rate thanks to a combination of smart timing, developer-focused innovation, and a bit of luck.
Nvidia Bets on Open AI Models to Secure Future in a Shifting Landscape
Nvidia is diversifying its AI strategy by releasing open models, data, and tools, aiming to maintain relevance as competitors like OpenAI and Chinese firms develop their own chips and accelerate AI innovation.
NeurIPS 2025: RL Reign, Google’s Surge, and a Party Boat of Ideas
NeurIPS 2025 saw reinforcement learning dominate discussions, alongside a surge in Google’s momentum and a vibrant, albeit chaotic, party scene. Attendees identified key trends, including the rise of continual learning and burgeoning interest in physical AI.
Nvidia's Data Center Boom: Sustainable or Tech Mania?
Nvidia's data center business has exploded to nearly $50 billion, fueled by AI infrastructure spending, but questions remain about the sustainability of this growth and whether it's driven by genuine AI potential or simply a speculative ‘bubble’.
Hugging Face CEO Warns of ‘LLM Bubble’ Bursting, Advocates for Specialized Models
Hugging Face co-founder Clem Delangue believes the current hype surrounding large language models is unsustainable, predicting a ‘bubble burst’ and advocating for a future dominated by smaller, specialized AI models tailored to specific industry needs.
Reflection Raises $2 Billion, Positioning Itself as an Open-Source Frontier AI Rival
Startup Reflection has secured a $2 billion valuation, fueled by its ambition to create an open-source alternative to closed AI labs like OpenAI and Anthropic, attracting backing from Nvidia and other major investors.
DeepSeek Unveils V3.2-exp: Sparsely Attentive Model Cuts Inference Costs
Researchers at DeepSeek have released V3.2-exp, a new experimental language model featuring DeepSeek Sparse Attention, designed to drastically reduce inference costs for long-context operations. The model utilizes a ‘lightning indexer’ and ‘fine-grained token selection system’ to prioritize context excerpts and load only relevant tokens into its attention window.
MicroFactory's Compact Robotic Workstation Offers a Novel Approach to Automation
San Francisco-based MicroFactory is developing a small, tabletop manufacturing kit utilizing robotic arms and AI, offering a potentially simpler and more accessible path to automation compared to traditional, human-sized robotic systems.
Belarusian Startup Vocal Image Sees Rapid Growth with AI-Powered Voice Coaching
Vocal Image, a Belarus-based startup focused on AI-powered voice and communication skills training, is experiencing rapid growth, boasting 4 million app downloads and a sizable user base. Driven by the founder's own journey and bolstered by recent funding rounds, the company is leveraging a vast, community-labeled voice dataset to improve its AI accuracy.
ByteDance Unveils Seed-OSS-36B: A New Open-Source LLM Challenge
TikTok’s parent company, ByteDance, has released Seed-OSS-36B, a new 36 billion parameter open-source large language model designed to challenge OpenAI and Anthropic, boasting enhanced reasoning capabilities and a longer token context.
AI Efficiency: Rethinking Compute to Reduce Waste
Experts are urging AI developers and enterprises to prioritize smarter model design and efficient computation over simply scaling up hardware, advocating for a shift in mindset towards reducing energy consumption and optimizing resource utilization.