A technical comparison of the NVIDIA H200 and AMD Instinct MI325X finds that the AMD chip's 256GB of HBM3e memory and 6.0 TB/s bandwidth let it host larger shards of 400B-parameter and 671B MoE models…
Aleph Alpha's Kolibri lost all 136 blind comparisons against Claude Sonnet 5.5 in a German-language Bible deep-dive benchmark run by Dewfall developer on 3 October 2026, scoring 3.7 with German prompt…
AI21 Labs cut high-priority AI job wait times from 72 hours to 12 hours — an 83% reduction — and reduced manual scheduling interventions from 20 per week to zero by adopting Google Cloud AI Hypercompu…
Aikido Security released Altar-1, an open-weight security model pruned from Z.AI's 753B-parameter GLM-5.3 MoE down to 328.0 GB of stored weights, a 78.2% reduction from BF16 and 32.8% smaller than its…
BottleCap AI released ThinkingCap-Qwen3.8-27B, a fine-tune of Qwen3.8-27B that cuts thinking tokens by an average of 37.2% across 12 benchmarks while macro-average accuracy falls 0.86 percentage point…
Cast AI's 2026 Kubernetes Optimization Report, analyzing tens of thousands of production clusters from January 2025 through April 2026, recorded average GPU compute utilization of 5% before any optimi…
Nunchux AI researchers Muyang Li and collaborators published the VC-Attention paper on September 14, reporting attention-kernel speedups of 1.46x to 1.59x on NVIDIA H200 and B200 data-center GPUs and …
A controlled pilot on a dedicated NVIDIA H200 GPU found ISO-AdamW scored 758 correct answers versus 754 for baseline AdamW on a 1,000-question held-out test, a +0.4% delta with a 95% confidence interv…
A benchmarking study by the aistack team found that the choice of coding agent harness has less impact than expected on task resolution accuracy, but can double the cost of a single resolved task from…
Anduril Industries, a defense technology company, is hiring a Senior AI Infrastructure Engineer in Costa Mesa, California, offering $166k–220k/yr, to lead GPU cluster robustness and automated resilien…
Signal65's new PINNACLE benchmark, developed with Kamiwaza, evaluates AI systems on whole enterprise task completion rather than model scores or token throughput, finding that 43 of 44 configurations …
A new study from arXiv (2608.23807v1) characterizes serving of masked diffusion language models (dLLMs) using LLaDA-8B-Instruct with a D2F LoRA adapter on a single NVIDIA H200 GPU, finding that reques…
IBM released two compact 470M-parameter English speech recognition models, granite-speech-5.0-470m-turboctc and granite-speech-5.0-470m-turboctc-nc, achieving over 12,600 RTFx on an NVIDIA H200 GPU an…
Apple's M5 Ultra Mac Studio, announced August 25, 2026, offers up to 512GB of unified memory at 1.2TB/s, enough to hold trillion-parameter-class open-weight models like DeepSeek-V4-Pro (1.6T params, ~…
NVIDIA introduced the ALCHEMI Toolkit, a PyTorch-native suite for GPU-accelerated atomistic simulations with machine learning interatomic potentials, and demonstrated that AI coding agents can build s…
A community release called Qwen3.8-27B-AEON-ULTIMATE-UNCENSORED-BF16, built on Alibaba's Qwen3.8-27B by the team AEON-7, removes the model's refusal behavior via abliteration and reports a mean KL div…
A technical analysis by an unnamed author, inspired by a post from Su Jianlin, derives the arithmetic intensity (FLOPs per byte) of attention during single-token decode, finding MHA has an AI of exact…
Mistral released the Mistral 3 family on December 2, 2025, under Apache 2.0, including nine dense Ministral 3 models (3B, 8B, 14B) and the sparse mixture-of-experts Mistral Large 3 with 675B total and…
Cast AI benchmark testing shows that continuous batching at batch size 8 reduces Llama 3.1 70B inference cost on a single H100 from approximately $0.60-$0.80 per million tokens to $0.15-$0.25 per mill…
Thinking Machines Lab released Inkling-Small, an open weights Mixture-of-Experts model with 276B total and 12B active parameters, about a quarter the size of its 975B-parameter Inkling model. The mode…