NVIDIA reported that combining its Transformer Engine library with the JAX Python library raised DeepSeek-V3 MoE training throughput on NVIDIA GB200 GPUs from an unoptimized baseline of 103 TFLOPS/GPU…
AllSpark released Iris-mini and Iris-pro, two open-weight search agents built on Qwen architectures and tuned for search and tool-use, which the team says lead their respective size classes in search-…
Apex Grid, a regtech developer building tools for Nigerian microfinance banks, has detailed its hybrid architecture that pairs vector databases with a versioned "data substrate" to make AI-driven comp…
The AllSpark team released Iris-mini and Iris-pro, two open-source search agents built on Qwen models that lead benchmarks among open-weight models in their size classes, according to the accompanying…
The developer of ZYR3.1MoE, an open-weights 9B-parameter multi-agent orchestration model built on a Qwen base, is seeking one or two collaborators for dataset curation, agent orchestration, or fine-tu…
Hetzner launched an experimental, free OpenAI-compatible Inference API in July 2026, serving the open-source Qwen3.6-35B-A3B-FP8 Mixture-of-Experts model from its data centers in Germany and Finland. …
Dify has emerged as the leading open-source orchestration framework for production LLM stacks moving from basic chain scripts to autonomous multi-agent systems, according to the article. Dify couples …
Qwen released Qwen-Drive-1.0-4B, a 4-billion-parameter vision-language model built on Qwen3.5-4B that fuses perception and planning into a single model with a bird's-eye-view perception head and two p…
NVIDIA Research's Efficient AI Team and Singapore Lab reported that a two-stage pipeline specialized for a single NVIDIA DGX Spark generates a 384p draft and refines it to 768p, producing a 5-second 7…
A data scientist's informal test found that six of seven major chatbots — ChatGPT, Gemini, Claude, DeepSeek, Llama and Qwen — all returned 27 when asked for a random number between 1 and 50, with only…
Anthropic's September 10, 2026 threat intelligence report documented approximately 200 million exchanges across five distillation campaigns by seven China-based labs, with Alibaba's Qwen team accounti…
Anthropic's September 2026 threat intelligence report documented five distillation campaigns that generated nearly 190 million exchanges with its Claude model between May and July 2026, with operators…
Anthropic's new threat intelligence report documents eight months of Claude abuse, including Chinese AI labs Alibaba's Qwen team, DeepSeek, and Moonshot AI relaying requests en masse or extracting tra…
Independent researcher Jinho Jang published a cross-scale mechanistic study of safety training in Mixture of Experts reasoning models, reporting 40+ novel findings from 200+ controlled experiments acr…
Anthropic said in a Thursday report that Alibaba, Moonshot, DeepSeek, Zhipu, and Xiaomi combined launched nearly 190 million distillation attacks against Claude between May and July, including 23 mill…
A production skill router covering 34,396 skills shows that fine-tuning on synthetic data improves in-distribution skill retrieval for LLM agents but causes catastrophic forgetting on real and out-of-…
ByteDance's Seed team, working with Singapore University of Technology and Design, Georgia Tech, and other institutions, released HarnessDev, a research project showing that LLMs can generate and iter…
Anthropic said on Thursday it disrupted China-based operators that used its Claude AI models for unauthorized model copying, surveillance, cyber intrusions, weapons-related research and fraud, accordi…
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…
DeepSeek released V4.1 Flash with a 1M context window, FP4 KV cache, and cross-layer attention reuse, while Alibaba open-weighted Qwen3.8-2.4T-A95B, a Qwen-Max-class mixture-of-experts model with 2.4 …