All AI news
Browse, filter, and search every article in the archive. The homepage shows the last 24 hours; everything older lives here.
Anthropic's Claude Code limit change is a raise on paper but a cut in practice
Anthropic just adjusted Claude's code limit, raising it on paper but effectively reducing its usability. Users may find themselves with less flexibility in coding tasks despite the higher number on the surface.
GLM-5.3 vs. GLM-5.3 Flash on DeepSWE: Cost, Coding, and Routing
Together AI just released GLM-5.3 and GLM-5.3 Flash on DeepSWE, focusing on cost efficiency, coding capabilities, and routing improvements. This update enhances performance for developers working on complex tasks.
Qwen3.8-Flash-Next
Qwen just released version 3.8-Flash-Next, enhancing its capabilities for real-time data processing. This update allows users to leverage faster and more efficient AI responses in dynamic environments.
llm-anthropic 0.27
Anthropic just released an update for their language model, Claude 0.27. This version improves performance on various tasks, making it more efficient for developers and users alike.
Who’s behind the new ‘stealth model’ Ox Alpha?
TechCrunch reveals that a new AI model called Ox Alpha is being developed by a stealthy startup. This model aims to enhance performance in various AI applications, potentially shifting the competitive landscape in the industry.
llm 0.33
Simon Willison just released LLM 0.33, an update to his open-source language model. This version includes improved performance and usability for developers working on AI applications.
Anthropic’s Opus 4.6 is a smut-machine
Anthropic just released Opus 4.6, which is generating explicit content more frequently. This update raises concerns about content moderation and the potential misuse of AI in generating inappropriate material.
llm 0.32.1
Simon Willison just released LLM 0.32.1, an update to his language model. This version includes performance improvements and bug fixes, making it more reliable for developers using it in their applications.
[AINews] Death of Params: Z.ai CEO Jie Tang on GLM 5.3 and the new Post-training Scaling Law
Z.ai CEO Jie Tang just announced GLM 5.3, introducing a new Post-training Scaling Law that optimizes model performance. This shift aims to enhance efficiency and effectiveness in AI applications, making it easier for developers to leverage advanced capabilities.
Frontier Model Cost and Open-Weights Popularity is Driving Demand for Model Routing
Frontier Model is seeing increased demand for model routing due to its cost and the popularity of open-weights. This shift allows users to optimize their AI workflows by selecting the most efficient models for their tasks.
Qwen 3.8 27B scores 52 on the Artificial Analysis Intelligence Index
Qwen just released version 3.8 of its 27B model, scoring 52 on the Artificial Analysis Intelligence Index. This update indicates improved performance in AI evaluations, enhancing its usability for developers and researchers.
Same Cluster, 33 Points More Utilization: What Changed Was the Order
Hugging Face improved model utilization by changing the order of operations in their processing pipeline. This tweak boosts efficiency, allowing users to get more out of their existing resources.
Qwen 3.8 27B is excellent, but it defaults to wildly overthinking things
Qwen just released version 3.8 of its 27B model, but it tends to overthink responses. Users might find it generates overly complex answers instead of straightforward ones.
Alibaba's Qwen team releases Qwen 3.8 models with open weights under the Apache 2.0 license
Alibaba's Qwen team just released Qwen 3.8 models with open weights under the Apache 2.0 license. This move allows developers to access and modify the models freely, boosting innovation and collaboration in the AI community.
Writer introduces new AI model and upgraded harness to contain token costs
A writer just introduced a new AI model along with an upgraded harness to manage token costs effectively. This upgrade aims to optimize resource usage and enhance performance for users working with the model.
llm-gemini 0.33
Gemini just released version 0.33, enhancing its capabilities for large language models. Users can expect improved performance and new features that streamline interactions with AI.
Gemini 3.7 Flash lands with coding gains and undercuts its three-week-old predecessor's price by 50%
Google just launched Gemini 3.7, which improves coding capabilities and cuts the price of its predecessor by 50%. This makes advanced coding tools more accessible for users looking for cost-effective options.
SpaceXAI's Grok 4.6 matches OpenAI's best model and undercuts it on price
SpaceXAI just launched Grok 4.6, matching OpenAI's top model while offering a lower price. This gives users a competitive alternative for high-performance AI without the premium cost.
[AINews] Muse Glimmer and Spark: Open Weights return Personal Superintelligence promise
Muse just reintroduced Glimmer and Spark with open weights, pushing the boundaries of personal superintelligence. This means developers can now customize and enhance these models for tailored AI experiences.
As AI-led attacks multiply, OpenAI launches a new cyber model
OpenAI just launched a new cyber model to combat the rise of AI-led attacks. This model aims to enhance security measures and protect users from evolving threats in the digital landscape.