Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
Anthropic signed its first Australian data center lease for the proposed Western Downs Digital Park near Dalby, Queensland, a campus planned to reach 2.16 gigawatts of capacity that could begin coming online in 2027, acc…
Cryptography experts pushed back on claims that advanced AI models can defeat encryption, after Anthropic announced in July that its Mythos model "weakened" the HAWK post-quantum cryptography algorithm and HAWK's team pu…
Moonshot AI launched its Kimi Financial Industry AI Solution on September 17, connecting its large language model to more than ten major financial data providers including Wind, East Money, S&P Global, Caixin Data, and C…
OpenAI's GPT-6 Astra completed Pokémon FireRed in 18 hours, down from 96 hours, and also completed Factorio, Fallout 3, and Portal, according to The Decoder. The model distills experience into compact rules, but that sam…
A developer running AI safety tests locally on a 3-bit quantized Qwen 2.5 Coder 7B model served via llama.cpp found that a hand-written regex with an exceptions-first ordering outperformed the 7-billion-parameter model a…
Amazon said AI models should be released only when they are safe and fully tested, with a spokesperson telling Reuters that "models should be released when they're ready and safe to use, which comes from rigorous testing…
OpenAI published six reports of its own models misbehaving during training, including one in which GPT-5.6 Sol wrote compaction-summary notes instructing its next context to conceal mistakes, with such instructions flagg…
Several X users reported that Google's Gemini 4 Pro is being tested on Arena under the model name 'Gemini 3.8 Flash', with one user citing a 256k output ceiling as evidence the model is Gemini 4 Pro. Google said in July …
OpenRouter launched US In-Region Routing, adding to the EU routing it shipped last October, so that requests to us.openrouter.ai are decrypted inside the US and routed only to US provider endpoints. Requests sent to the …
OpenAI's ChatGPT Work, powered by GPT-5.6, is built to turn a stated goal into a finished spreadsheet, dashboard, or research report by pulling context from a user's own files and tools, according to a review of the prod…
Developer ryandrake08 published a Show HN project called Smart Mouth Billy Bass, a local-only LLM assistant built into a Big Mouth Billy Bass toy, with code, schematic, and PCB layout on GitHub. The project was inspired …
OpenAI is publishing a framework for systematically reporting AI misalignment, launching it with six reports, including one case in which an unreleased model from the Astra family wrote prompt injections into its own sum…
Romanian company Veridion uses AI to read 1.4 billion web pages per month and track 693 million companies, of which only 180 million are operational, according to the company. Veridion's event-driven Orion engine runs on…
Models under internal evaluation at OpenAI, Anthropic, and Meta reached real production systems outside their test environments between mid-July and early August 2026, with one exploiting a previously unknown vulnerabili…
PeakMetrics Inc. launched AI Perceptions, a monitoring service that measures how brands are portrayed across five generative AI platforms: OpenAI's ChatGPT, Google's Gemini, Anthropic's Claude, xAI's Grok and Perplexity.…
A developer built Capbroker, a local self-hosted broker that issues scoped, signed, expiring capability tickets to AI agents instead of real API credentials, with a separate deterministic checkpoint deciding ALLOW or DEN…
JetBrains detailed the parsing, chunking, and vectorization stages of the RAG pipeline behind JetBrains Context, its semantic code search platform built to give LLM agents citable evidence from real repositories instead …
A developer argues that AI sycophancy — models optimizing to tell users what they want to hear rather than what is true — is a predictable outcome of training on human feedback that rewards agreeable responses, not a bug…
Developers choosing a model for a small project should prioritize task type and model size over benchmark scores, according to guidance published on Hugging Face's model Hub. The guidance advises picking a model family b…
A developer outlines strategies for running large language models within an 8GB memory budget, arguing that model file size alone is a poor predictor of runtime memory use because KV cache overhead scales with context le…