Granite 4.0 3B Vision: compact multimodal intelligence for enterprise documents
The IBM Granite team released the Granite 4.0 3B Vision model, a compact multimodal large language model designed for…
AI news monthly archive: March 2026 features 171 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.
The IBM Granite team released the Granite 4.0 3B Vision model, a compact multimodal large language model designed for…
OpenAI has raised $122 billion in new funding to expand frontier AI globally and invest in next-generation compute,…
Trail of Bits shares its systematic approach to moving AI adoption from about 5% to company-wide implementation within…
KwaiKAT released the non-reasoning code model KAT-Coder-Pro V2, which scored 44 on the Artificial Analysis…
The OpenMed team built an end-to-end AI pipeline covering protein structure prediction, sequence design, and codon…
Agent Development Kit (ADK) for Go 1.0 has been officially released, marking its transition from experimental scripts…
Orbax and MaxText introduced a new continuous checkpointing feature designed to optimize the balance between…
Meta released a new AI model called Bayesian Optimization for designing concrete mixes. The model aims to help the…
This is a very good post: [Quoting @boazbaraktcs]: New blog post: The state of AI safety, four fake images.…
Introducing Critique, a new multi-model deep research system in M365 Copilot. You can use multiple models…
Inception Labs released Mercury Edit 2, a diffusion LLM (dLLM) built specifically for next-edit prediction,…
The Meituan LongCat team's open-source diffusion TTS model abandons the traditional mel-spectrogram intermediate…
The author plans to introduce a series of hidden and underused features in Claude Code, focusing on the ones used most…
Google officially released version 1.0.0 of the Java Agent Development Kit (ADK). The release introduces several key…
Claude raised the max_tokens limit for Opus 4.6 and Sonnet 4.6 in the Message Batches API to 300k. The…
Qwen Studio has been released, integrating chatbot, image and video understanding, image generation, document…
The author uses a hypothetical scenario involving the game "Glurg" (actually chess) to argue that, with existing…
Current frontier multimodal large models can achieve top rankings on standard chest X-ray question-answering…
The poster first spent hours using an LLM to carefully refine a blog argument, then asked it to argue the opposite…
OpenAI launched the Codex use cases library, covering practical scenarios for coding and non-coding tasks. Users can…
A receptionist demo built with gpt-realtime-1.5 for a Singapore clinic holds natural conversations with patients,…
OpenAI Codex launched a library of plugin use cases covering scenarios such as building iOS apps, analyzing datasets…
This report presents the training process of the code model Composer 2. The model is based on the open-source…
Codex has reset usage limits across all plans so users can try the newly launched plugins. You can now build projects…
Codex officially launched Plugins, with out-of-the-box support for mainstream development tools such as Slack, Figma,…
Codex has officially launched its plugin system, with out-of-the-box support for Slack, Figma, Notion, Gmail and other…
Google released Gemini 3.1 Flash Live, calling it its highest-quality audio model to date, with lower latency, higher…
The biggest challenge in building modern applications is not the code itself but the tedious service integration, API…
The Gemini 3.1 Flash Live audio model has been released, supporting more natural real-time conversations and improved…
Gemini Live has been upgraded to run on Gemini 3.1 Flash Live, with faster responses and fewer pauses, support for…
Gemini 3.1 Flash Live has been released as Google's highest-quality audio and speech model to date, with significantly…
Gemini 3.1 Flash Live is now available across Google's product lineup, delivering more natural and reliable real-time…
OpenAI released the lightweight GPT-5.4 mini and nano models, retaining multiple reasoning levels and a 400K context…
The Cursor team applied real-time reinforcement learning to the Composer coding model, using reasoning tokens…
Anthropic's engineering blog published a post on the design of Claude Code auto mode. The mode serves as a safe middle…
The UK AI Security Institute, in partnership with the Bank of England, monitored 177,436 public MCP server tools…
Sakana AI, together with the University of British Columbia, the Vector Institute, and the University of Oxford,…
Google DeepMind launched Lyria 3 Pro, which generates up to 3 minutes of high-fidelity music with free arrangement of…
Claude mobile now integrates work tools, allowing users to view Figma designs, create Canva slides, and browse…
LLM memory features suffer from over-association: a casual question two months ago gets permanently archived as a deep…
Page 1 of 5
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