Models & Research

DeepSeek just released V4 — two open-source models under MIT license that come dangerously close to frontier performance. Pro has 1.6 trillion parameters, Flash costs almost nothing.

Google's April Drop brings NotebookLM directly into the Gemini app, launches a native Mac app, and lets Gemini read your emails for better meeting prep.

Google Cloud chief Thomas Kurian confirmed that Gemini will power a new, personalized version of Apple's Siri later in 2026. This is huge.

Just six weeks after GPT-5.4, OpenAI ships the next model. GPT-5.5 'Spud' promises to handle complex tasks with less hand-holding — and uses fewer tokens.

Alibaba's new open-source model needs just 18 GB of RAM yet outperforms its own 397 billion parameter flagship on coding benchmarks. Apache 2.0.

OpenAI launches gpt-image-2, an image model that reasons through prompts before generating. 2K resolution, better text rendering, and multi-image consistency included.

Google is bringing its AI assistant in Chrome to Australia, Indonesia, Japan and four more Asian markets. Browser AI goes global.

Dario Amodei met with Trump's Chief of Staff Susie Wiles and Treasury Secretary Bessent on Friday. After months of icy relations over the Pentagon dispute, a thaw is emerging.

Nine Claude agents solved an open alignment problem better than Anthropic's human researchers. The implications are far-reaching.

Anthropic will retire Claude Haiku 3 on April 20. API calls will return errors after that. The successor Haiku 4.5 costs four times as much.

AI pioneer Yoshua Bengio is calling for an FDA for AI. The reason: Anthropic's Mythos dominated the IMF Spring Meetings and exposed how much power a single company holds over global infrastructure.

GPT-Rosalind is OpenAI's first specialized model for life sciences. Access is limited to select partners like Moderna and Amgen.

Google released a text-to-speech model you can control with prompts. From 'enthusiastic' to 'Brixton accent' — it's all there.

Google connects Nano Banana to your photo library. Say 'make a picture of me as an astronaut' — and Gemini knows what you look like.

Anthropic officially launches Claude Opus 4.7. The new flagship model brings 13% better coding performance, triple the image processing capacity, and a new 'xhigh' effort level.

Google turns Gemini prompts in Chrome into saveable 'Skills' - create once, run anywhere with a single click. A clever feature that shows where browser AI is heading.

Google launches Gemini as a standalone desktop app for Mac. Hit Option+Space and you've got AI anywhere — the three-way desktop battle just got real.

The best AI agents achieve only half the performance of PhD experts on complex tasks. A Nature study pumps the brakes on the agent hype - while delivering some surprising nuances.

OpenAI's next model codenamed 'Spud' has finished pre-training. Polymarket gives it 78 percent odds of launching by end of April.

OpenAI releases a specialized cybersecurity model. Their response to Anthropic's Mythos came faster than expected.