N8n has added a dedicated Qwen Cloud node, integrating Alibaba Cloud's Qwen model family into its workflow automation canvas with support for text, image, and video actions. The node, available in n8n…
A new method called Functional Reconstruction improves draft-token acceptance in speculative decoding for multi-head latent attention (MLA) draft models, according to a paper on arXiv (2607.27269v1). …
Every notable open-weight frontier model released in the past year uses a Mixture-of-Experts (MoE) architecture, according to an analysis by Vetted Consumer. The shift means total parameters range fro…
ElevenLabs has introduced Expressive Mode, a feature within its ElevenLabs Agents platform that adds emotional tone and inflection control to real-time voice agents, allowing them to sound annoyed, am…
A developer shared a configuration guide for integrating Qwen AI models into OpenCode on Ubuntu Linux, providing the JSON configuration file path and code snippet for setting up the Qwen provider with…
Frontier AI labs are building subtle developer lock-in by shifting from model weights to proprietary state and execution infrastructure, OpenAI's GPT-5.6 Sol achieved a 38.3% ARC-AGI-3 score (up from …
Researchers have developed Acoda, a genetic algorithm-based adversarial code obfuscation framework that defends against LLM-based code analysis by inducing LLMs to refuse or misinterpret code. On seve…
Unit 42 identified a Chinese-speaking threat actor using the Hermes Agent framework with DeepSeek as an autonomous offensive operator to target infrastructure via seven vulnerabilities, combining AI-d…
A developer built HOUJIAYAN API, an OpenAI-compatible gateway that routes requests across 16 Chinese LLMs including DeepSeek, Kimi, Qwen, MiniMax, and Doubao. The gateway handles authentication, billi…
Researchers introduced a model-side aggregation interface using Hash-based HyperLogLog (HLL) sketches that maintains compact 2 KiB states alongside frozen language models, achieving 1.6% mean relative…
Perihelion, a 13 MB Rust-based coding agent compatible with Claude Code plugins and skills, launches with support for multiple API providers including DeepSeek, GLM, Qwen, and Anthropic, achieving 98%…
Anthropic, the only major frontier AI lab that refused to sign an Nvidia-led open letter defending open-weight AI models, now argues it opposes a ban on such models while advocating restrictions that …
Google Cloud will retire Moonshot AI's Kimi K2 Thinking model endpoint on October 21st, after deprecating it on July 21st as part of a broader endpoint retirement that includes models from DeepSeek, Z…
Chinese students like 18-year-old Guo Xinyan are increasingly using AI chatbots from Alibaba, ByteDance, Tencent, and Baidu to help select colleges and majors after the gaokao exam, as families seek f…
OpenCode, an open-source coding agent, can now run entirely locally using QVAC, an SDK and CLI from Tether that serves models on-device across GPU backends. The setup eliminates cloud dependencies, gi…
China's AI advantage lies in affordability and accessibility rather than peak performance, with Chinese models processing 18 trillion tokens weekly on OpenRouter by June, triple the U.S. total, accord…
Epilude released Model 4.1, a refined on-device cleanup model for Local Mode on Mac, that reduces critical errors from 4 to 3 and passes 79 of 90 internal benchmark scenarios, up from 78. The model is…
A developer stretched Qwen 35B's context window from 4,096 to 32,768 tokens on a 12GB RTX 4070 by quantizing the KV cache to q8_0 using llama.cpp flags -ctk q8_0 -ctv q8_0, achieving an 8× context inc…
A study of 45 language models found that appending a two-word confirmation tag like 'right?' to a decision question can shift model agreement by up to 64 percentage points, with newer models showing i…
Fermisense, an AI consultancy, reports that a 9B Qwen specialist fine-tuned for $500 on two RTX PRO 6000 GPUs over three and a half days scored 87.3% on a simulated catalog-review benchmark, outperfor…