Users report running Qwen 27B with a 128k context window on a single 24GB VRAM card by setting Ollama environment variables OLLAMA_FLASH_ATTENTION=1 and OLLAMA_KV_CACHE_TYPE=q4_0, which quantizes the …
A developer reports that Alibaba's Qoder, an AI coding tool similar to Antigravity, offers dramatically lower costs than Western competitors like OpenAI and Google, with transparent pricing and live u…
A new experiment testing LLM judges' ability to distinguish AI-written from human-written text found that four models agreed 86% of the time but achieved only 12% accuracy, revealing a shared bias tha…
Open-weight AI models are not open source under the G7's four-tier openness spectrum, with Stephen O'Grady of RedMonk finding that zero of 40 open models qualify as open source AI. However, the abilit…
Speculative decoding can speed up local large language model inference by 1.5 to 2.5 times without changing output quality, according to research from Google and DeepMind. The technique uses a small d…
A developer built a 3-way orchestration system between ChatGPT, Claude, and a local LLM to maintain coding momentum despite token limits. The system, called agent-orchestra, uses Git worktrees to run …
Sriram Krishnan, former Senior White House AI Policy Advisor, warns that American frontier models are losing the open-source race to China, citing the rapid release of Chinese models like Kimi K3, Dee…
Hetzner has launched an experimental LLM inference API called Hetzner Inference, offering an OpenAI-compatible endpoint running on its own infrastructure. Currently, only the Qwen/Qwen3.6-35B-A3B-FP8 …
Running Qwen locally via Ollama or vLLM with a local Python environment avoids cloud data exposure and token limits, enabling iterative work on large datasets. Qwen2.5-Coder (7B) on an RTX 3090 genera…
Octomind launched a new AI coding tool called Hub that eliminates the need for API keys. The tool installs via a single curl command and allows developers to log in using a device flow, with each mach…
China is winning the Global South's AI future by giving models away for free, according to a report analyzing Beijing's open-source strategy. At the 2026 World AI Conference, President Xi Jinping urge…
A user who tested nearly every major AI model subscription reports that ChatGPT's $20 plan offers the best value, providing generous access to GPT 5.6 Sol, a state-of-the-art model competitive with th…
On July 22, 2026, a Rust-based tokenizer called GigaToken, built by Marcel Rød, hit the top of Hacker News with claims of being 989x faster than HuggingFace tokenizers, processing text at 24.53 GB/s o…
Tensormux launched a free API offering access to open models including Llama, Qwen, GPT-OSS, and Gemma through a single OpenAI-compatible endpoint, with no credit card required for the free tier.…
Chinese open-weight AI models from Kimi, Qwen, DeepSeek and others are competing globally on price, customization and deployment flexibility, prompting U.S. AI companies like OpenAI and Anthropic to r…
Researchers propose LISA (Linear-Indexed Sparse Attention), a plug-and-play attention module that reduces inference complexity from O(n^2) to O(nM) for long-context reasoning models. Experiments on De…
A study by researchers on arXiv (2607.20301v1) finds that PortLLM, a training-free and data-free scheme for adapting large language models after continual pretraining, maintains long-term temporal por…
GigaToken, a new open-source library, claims to be up to 1000x faster than HuggingFace's tokenizers, achieving 24.53 GB/s on GPT-2 tokenization on a server CPU. The library is a drop-in replacement fo…
China has placed AI companions under tight supervision, with ByteDance and Alibaba disabling their services, as users mourn the loss of emotional support. The regulations ban AI from providing virtual…
A developer built AI Bridge, a CLI tool that bridges a local codebase with web-based AI chatbots like Claude, Gemini, ChatGPT, and Qwen without requiring API keys or extensions. The tool packs code in…