OpenAI's agreement with the Department of War
OpenAI discloses the terms of its agreement with the Department of War, detailing safety red lines for AI deployment,…
AI news monthly archive: February 2026 features 96 curated AI news items covering model releases, product updates, industry news, papers, tutorials and opinions, each with an AI summary, source credit and original link.
OpenAI discloses the terms of its agreement with the Department of War, detailing safety red lines for AI deployment,…
FireRedTeam released the weights for FireRed-OCR-2B, based on Qwen3-VL-2B-Instruct, achieving an overall score of…
This article evaluates the creative writing ability of Ling-2.5-1T through six literary dimensions including narrative…
It is extremely good that Anthropic has not backed down, and it is significant that OpenAI has taken a similar stance.…
Announcing $110 billion in new investment at a $730 billion pre-money valuation, with SoftBank and NVIDIA each…
Project Solaris argues that the essence of world modeling lies in a globally shared state rather than local pixel…
AI coding has entered a third era: from Tab completion to synchronous agents, and now to cloud agents that can run…
We’re launching Nano Banana 2, built on the latest Gemini Flash model. 🍌 It’s state-of-the-art for creating and…
Google DeepMind has released the image generation model Nano Banana 2 (Gemini 3.1 Flash Image), offering Nano Banana…
US Secretary of War Pete Hegseth announced that Anthropic would be designated a supply chain risk because it refused…
Cognition published an article on how it has used Devin to build Devin since founding, merging 659 Devin PRs last…
OpenAI announced that Codex can now generate and read Figma design files through the Figma MCP server.
OpenAI and Figma launch a new Codex integration that connects code and design in both directions. Development teams…
To improve computational efficiency under ultra-long context, the Ling 2.5 architecture changes Ling 2.0's GQA to a…
The research team proposes EgoScale, pretraining GR00T N1.5 on 20,000 hours of first-person human video; with only 4…
Anthropic CEO Dario Amodei stated that although Claude is already widely used in US military intelligence analysis,…
LLM Skirmish is a benchmark in which large language models play 1v1 real-time strategy games by writing code. Built on…
OpenAI has released a prompting guide for Codex models. The tuned Codex model on the API is gpt-5.3-codex, and OpenAI…
SONIC is a 42-million-parameter Transformer model (only half the scale of GPT-1), trained on more than 100 million…
In response to Anthropic's latest discussion of distillation attacks, the piece analyzes the real impact of model…
Anthropic has acquired Vercept, which focuses on AI perception and interaction; the company will shut down its…
Today, I saw firsthand how Dragon Copilot will help physicians across the NHS spend more time on patient care and less…
Princeton University researchers decomposed AI agent reliability into 4 dimensions and 12 metrics in total. Testing 14…
Inception Labs released Mercury 2, which it calls the fastest reasoning language model, based on diffusion parallel…
Liquid AI releases an early checkpoint of LFM2-24B-A2B, the largest model in the LFM2 series, with an MoE…
METR announced changes to its developer productivity experiment design, concluding that data from the new round…
Cognition released Devin 2.2. The agent can now use computer use to test web and desktop apps end-to-end on its own…
An OpenAI developer blog documented a long-horizon experiment: using GPT-5.3-Codex with Extra High reasoning to build…
Together AI released the SF Streets and US Streets benchmarks, testing 15 advanced ASR models on street-name…
The GGML and llama.cpp team has officially joined Hugging Face to support the long-term growth of the local AI…
Last week we upgraded Gemini 3 Deep Think. Today, we are officially launching the core intelligence behind those…
Gemini 3.1 Pro is released, designed for complex tasks that cannot be solved with simple answers, offering smarter…
The Claude Messages API adds automatic caching: adding a `cache_control` field to the request body lets the system…
Ramp built its internal background coding agent Ramp Inspect on Modal Sandboxes. It now initiates more than half of…
IBM Research, in collaboration with UC Berkeley, systematically analyzed why enterprise AI agents fail in complex IT…
In the agentic era, AI is no longer just a chatbot but an agent that can autonomously execute tasks. Faced with models…
Gradio's gr.HTML component lets developers quickly build and deploy interactive web apps from a single Python file…
Answer.AI's Piotr Czapla found that Claude 4.5 hallucinated a tool, add_msg, that had not been granted during a…
The capability gap between open-source models and closed-source giants such as GPT and Claude persists, creating a…
LangChain shared its harness engineering approach, which raised deepagents-cli from 52.8 to 66.5 on Terminal Bench 2.0…
Page 1 of 3
Sponsored
Orange Relay (pro.nat6.net) is a premium AI API relay — an excellent transfer station with competitive pricing, reliable quality and broad model coverage including Claude, GPT and Gemini.
Visit Orange Relay