ARCQuant, a new framework for Large Language Model inference, uses the NVFP4 numerical format to achieve up to 3x speedup on GPUs while maintaining accuracy comparable to full-precision baselines. The…
Mini PCs with unified memory architectures, such as those using AMD's Strix Halo chip, can run 70-billion-parameter language models that exceed the VRAM capacity of discrete GPUs like the RTX 5090, bu…
A solo developer built Prometheus, an autonomous research system that runs 24/7 on a single Linux workstation with one RTX 5090 GPU, generating its own questions and running over 130,000 experiments. …
NVIDIA is preparing the GeForce RTX 5090 SE graphics card, a cut-down version of the flagship RTX 5090 based on the same GB202 silicon but with fewer cores and a 500W TGP, targeting a $1,500 MSRP. Unl…
Glad Labs fixed a GPU pinning issue where LiteLLM 1.89.2's global api_base override prevented per-model routing, causing vision tasks to cold-load onto the wrong GPU. The team also hardened content gu…
A Hacker News user asks the community for advice on moving from cloud-based AI services to private models for sensitive data analysis and reasoning, considering options like adding dual RTX 5090 GPUs …
Researchers introduced Hierarchical Global Attention (HGA), a method that reduces GPU memory usage in long-context transformers by using hierarchical routing, enabling a 64K-token context on an RTX 50…
Ahmad Osman, founder of Osmantic, argued at the AI Engineer World's Fair that local AI is rapidly catching up to proprietary frontier models, driven by shrinking gaps in open-source LLMs and improved …
GLM, the model family from Z.ai (formerly Zhipu AI), is gaining attention for its open weights and strong benchmark numbers. A developer tested GLM-4.7-Flash, a 30B-A3B MoE model, on a single-GPU home…
Glad Labs has developed an AI-operated content pipeline that treats content generation as an engineering problem, moving beyond simple prompting to autonomous agents with human oversight. The system u…
A new AI content pipeline called Poindexter uses agent infrastructure, RAG pipelines, and open-source LLMs to automate technical publishing, reducing the human loop and enabling solo developers to pro…
A technical analysis compares Mac hardware, Nvidia RTX 5090, and cloud services for running local large language models in 2026, warning consumers about spec-sheet traps that waste money.…
Gigabyte's $5,300 AORUS RTX 5090 Infinity, a limited-edition anniversary graphics card, is now available for in-store pickup at Micro Center in the US, but buyers will not receive the 1-gram 999 pure …
Amazon has dropped the Corsair Vengeance i5200 gaming PC featuring an RTX 5090 to its lowest price ever at $5,899.99, a $500 discount. The prebuilt system includes an Intel Core Ultra 9 285K, 64GB DDR…
NVIDIA Research released cuTile Rust, a tile-based system for writing memory-safe, data-race-free GPU kernels in Rust. The project extends Rust's ownership model to GPU programming, achieving up to 92…
Georgi Gerganov, creator of llama.cpp, reported that the Qwen3.6-27B model is highly capable for local coding tasks, using it daily on his M2 Ultra and RTX 5090 systems. He noted the model's utility f…
Newegg is offering a $845 discount on the ABS Kaze II Ruby gaming PC, which features an RTX 5090, Ryzen 7 9800X3D, 64GB DDR5, and 2TB SSD, bringing the price to $4,654.99. The deal includes a $600 ins…
Google DeepMind released DiffusionGemma, an open-source Apache 2.0-licensed diffusion-based large language model that generates text at over 1,000 tokens per second on a single H100 GPU, achieving up …
Amazon is offering the HP Omen Max 45L gaming desktop with an RTX 5090 graphics card and Ryzen 9 9900X3D processor for $600 off its regular price. The prebuilt system includes 32GB of DDR5 RAM, a 2TB …
NVIDIA Research and Tencent YoutuResearch have released open-source systems that replace the VAE decoder bottleneck in latent diffusion models, enabling faster 4K image generation. NVIDIA's PiD reform…