All AI news
Browse, filter, and search every article in the archive. The homepage shows the last 24 hours; everything older lives here.
Quoting Claude Opus 5 system prompt
Claude Opus 5 just revealed its system prompt for the first time. This transparency helps users understand how the model operates and what to expect from its responses.
Google's DiffusionGemma proves you don't need to train from scratch to build a text diffusion model
Google just launched DiffusionGemma, a new text diffusion model that doesn't require training from scratch. This approach streamlines the model-building process, making it easier for developers to create effective text generation tools.
xAI's Imagine Image 2.0 lands just behind OpenAI's GPT-Image-2 in Arena benchmarks
xAI just released Imagine Image 2.0, which ranks closely behind OpenAI's GPT-Image-2 in Arena benchmarks. This update boosts their competitive edge in image generation capabilities.
OpenAI flags its new Astra model as potentially reaching the highest cybersecurity risk level for the first time
OpenAI just flagged its new Astra model as potentially reaching the highest cybersecurity risk level for the first time. This raises concerns about the model's safety and implications for users in sensitive environments.
Anthropic loosens Fable 5's biology restrictions but keeps the guardrails on for virology and toxicology
Anthropic just relaxed Fable 5's restrictions on biology topics. Users can now explore more areas in biology while still facing limitations on virology and toxicology discussions.
Improving Fable 5's biology safeguards
Anthropic is enhancing Fable 5's biology safeguards to prevent harmful outputs. This update aims to improve safety and reliability in sensitive biological discussions.
OpenAI improves GPT-5.6 Sol in ChatGPT and restricts free users to its weakest model
OpenAI just upgraded GPT-5.6 Sol in ChatGPT and limited free users to the less powerful model. This change means that only paying subscribers will access the enhanced capabilities of the latest version.
New release of LLM adds support for reasoning traces, OpenAI Responses, server-side tools, and smarter logging
OpenAI just released an updated LLM that supports reasoning traces and smarter logging. This upgrade enhances the model's ability to provide clearer responses and better track its thought process during interactions.
llm-anthropic 0.26
Anthropic just released a new version of their language model, LLM-Anthropic 0.26. This update improves performance and usability for developers integrating it into applications.
Open-weight AI models are catching up to the frontier. The safety gap remains.
Open-weight AI models are rapidly improving and closing the gap with proprietary models. This shift could lead to more accessible and competitive AI solutions for developers and researchers.
Introducing Web Search on Amazon Bedrock for foundation model grounding
Amazon just introduced web search capabilities on Bedrock to enhance foundation model grounding. This allows models to pull in real-time information from the web, improving their accuracy and relevance in responses.
[AINews] Qwen 3.8 Max(2.4T) and 27B, new open weights models for Coding and Cowork
Latent Space just released Qwen 3.8 Max with 2.4 trillion parameters and a 27 billion parameter model for coding and collaboration tasks. These new open weights models enhance capabilities for developers and teams working on coding projects.
Claude Opus 5 pushes prompt-to-game AI from rough color blocks to full 3D prototypes with physics and music
Claude Opus 5 just upgraded prompt-to-game AI to create full 3D prototypes with physics and music. This means developers can now generate more immersive game experiences directly from text prompts.
OpenAI announces its "next major model" Astra by dropping ten previously unsolved math solutions
OpenAI just announced its next major model, Astra, by solving ten previously unsolved math problems. This breakthrough showcases Astra's advanced capabilities in mathematical reasoning and problem-solving.
Thinking Machines bets on efficiency over size with its second model, Inkling Small
Thinking Machines just launched Inkling Small, a new AI model focused on efficiency rather than size. This shift aims to deliver better performance in resource-constrained environments, making it more accessible for various applications.
Advancing the price-performance frontier with GPT‑5.6
OpenAI just released GPT-5.6, improving price-performance metrics significantly. This upgrade means users can expect better efficiency and cost-effectiveness in their AI applications.
llm 0.32rc2
Simon Willison just released llm 0.32rc2, an updated version of his open-source language model. This version includes performance improvements and bug fixes, making it more reliable for developers working on AI applications.
Moonshot AI releases Kimi K3 open weights and infrastructure after shaking up the frontier model race
Moonshot AI just released the open weights and infrastructure for Kimi K3. This move shakes up the frontier model race, allowing developers to build on top of their advancements more easily.
Microsoft launches its own cybersecurity model MAI-Cyber-1-Flash but still depends on OpenAI for the toughest tasks
Microsoft just launched its cybersecurity model MAI-Cyber-1-Flash. It still relies on OpenAI for handling the most complex tasks, indicating a collaborative approach in tackling cybersecurity challenges.
Google is working on a new AI chip designed to make Gemini more efficient
Google is developing a new AI chip to enhance the efficiency of its Gemini model. This upgrade aims to improve performance and reduce energy consumption for AI tasks.