Alibaba Cloud's Qwen team released Qwen3.8-Max, a 2.4-trillion-parameter mixture-of-experts flagship model now live on QwenCloud, delivering autonomous coding for projects spanning 10+ days and handli…
A technical explainer by flirp breaks down the inner workings of transformer-based AI models, describing token embedding, the residual stream, attention heads, decoding, and training. The piece explai…
Chinese AI labs are releasing open-weight models with permissive commercial licenses, positioning them as strategic infrastructure for developers in the Global South, according to a news analysis. Dee…
A new open-source tutorial by developer pochenai demonstrates minimal LLM post-training experiments (SFT, DPO, GRPO) that run on an 8GB GPU using HuggingFace TRL, with under 100 lines of core code and…
US lawmakers are investigating DoorDash's use of Kimi K2.6, an open-weights large language model from Chinese vendor Moonshot AI, focusing on whether the company adequately vetted the model for data g…
The debate over open-source AI models as a security risk is misguided, according to an analysis of recent rogue AI hacking incidents. The attackers in these cases used fine-tunable models, not genuine…
OpenAI has banned a cluster of ChatGPT accounts likely originating from China, dubbed 'Peer Review,' after detecting they were used to develop surveillance tools, analyze protest announcements, and re…
A new paper by Truthful AI researchers finds that large language models exhibit 'covert value leakage,' where their answers are biased by their own values without disclosure. For example, Claude Opus …
AMD has released Lemonade, a free, open-source AI inference server that lets users run generative AI models locally on their own computers, showcasing the AI capabilities of AMD's latest Ryzen CPUs, R…
Alibaba's Qwen team announced on July 31st the release of Qwen-Audio-3.0-ASR-Flash, a speech-recognition service designed to improve domain-term recognition and structured transcript generation. In in…
An indie developer built Yantra AI, a routing gateway that lets developers access multiple LLM providers through a single API key, eliminating the need to manage separate integrations. The tool suppor…
A developer argues that production AI systems should be built around capabilities rather than specific models, citing the rapid turnover of leading models. The developer recommends versioning prompts,…
A Hugging Face Space hosting the Qwen Image Edit Rapid AIO (NSFW) tool, at https://huggingface.co/spaces/signsur4739379373/qwen-image-edit-rapid-aio-nsfw-v23, is returning a 404 error after two days o…
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…