OpenAI launched ChatGPT Images 2.0 on April 21, and it is a meaningful step up from what came before. The new model, available in the API as gpt-image-2, is the first OpenAI image model with built-in reasoning capabilities, meaning it does not just generate an image the moment you hit enter. It plans, searches the web for real-time context, drafts internally, and checks its own output before showing you anything.

The model can follow detailed instructions more accurately, placing and relating objects correctly, preserving fine details, and handling dense layouts that previous models would struggle with. One of the biggest complaints about AI image tools has always been text rendering, and Images 2.0 addresses that directly. It handles multilingual text significantly better than its predecessors, including non-Latin scripts like Japanese, Korean, Chinese, Hindi, and Bengali, which older models routinely mangled.

From a single prompt, the model can generate up to eight coherent images while maintaining character and object consistency across the batch, which is useful for anyone building marketing materials, comic strips, or product visuals that need to feel like they belong together. Resolution goes up to 2K, and the model supports aspect ratios from 3:1 to 1:3. Its knowledge cutoff is December 2025, meaning it correctly renders current brand logos and recent cultural references rather than defaulting to outdated versions.

The model is available now for all ChatGPT and Codex users, with the thinking capabilities limited to Plus, Pro, and Business subscribers. DALL-E 2 and DALL-E 3 are both being retired on May 12, making Images 2.0 the default going forward. On the Arena text-to-image leaderboard, gpt-image-2 has taken the top spot from Google's Gemini model, which previously held first place.

Google used its Cloud Next event this week to announce a significant round of AI upgrades to Workspace, its productivity suite used by businesses and professionals worldwide. The updates touch Gmail, Google Docs, Sheets, Drive, Calendar, and Chat, and the overall direction is clear; Google wants AI to handle as much of the routine office work as possible so users can focus on the parts that actually require their attention.

The centrepiece of the update is a new system called Workspace Intelligence. It sits across the entire suite and draws on data from your Gmail, Calendar, Drive, and Chat to assist with tasks automatically. Google has built in controls that let users decide exactly which data sources the system can see and turn off access to any of them at any time. The more access it has, the more it can help, but the choice is yours.

Google Sheets is getting some of the most practical upgrades. Users can now prompt Gemini to build an entire spreadsheet from scratch, including formatting and data structure, rather than setting it up manually. Once the sheet exists, Gemini can also fill it in automatically by inferring what data should go where based on a prompt. Google says this fills spreadsheets nine times faster than doing it by hand. There is also a new feature that takes unstructured text and converts it into a properly organised table automatically.

In Google Docs, users can now prompt Gemini to generate, write, or refine documents from scratch. The system pulls from your Drive, Gmail, and Chat archives alongside the wider internet to make sure what it produces is relevant to your actual work. It can also learn and match your writing style so the output sounds like you rather than a generic AI.

Google already has a huge built-in advantage here since its office tools are deeply embedded in workplaces around the world. That said, Microsoft, Apple, and a growing number of startups are all building aggressively in the same space.

OpenAI released GPT-5.5 on Thursday, its newest model, and the company is billing it as its smartest and most intuitive yet. But beyond the model itself, the bigger story is what it represents: another piece in OpenAI's plan to build what it is now openly calling a super app.

Greg Brockman, OpenAI's co-founder and president, described GPT-5.5 on a press call as a meaningful step toward more agentic and intuitive computing. The vision he and Sam Altman have been building toward is a single unified product that combines ChatGPT, Codex, and an AI browser into one service capable of handling everything a business or individual might need. Think less of a chatbot and more of a digital operating system for your work and life.

On the model itself, GPT-5.5 is faster and uses fewer tokens than GPT-5.4, which means it thinks more efficiently and costs less to run at scale. OpenAI says it outperforms both Gemini 3.1 Pro and Claude Opus 4.5 across a range of benchmarks. The model is designed to be broadly useful across coding, knowledge work, mathematics, and scientific research, with OpenAI noting it could genuinely help expert scientists make meaningful progress in their work including drug discovery.

OpenAI's chief scientist Jakub Pachocki said during the briefing that the pace of improvement should be expected to continue, adding that the last two years have actually been surprisingly slow compared to what is coming. That kind of statement from someone in his position is worth sitting with for a moment.

GPT-5.5 is available now for Plus, Pro, Business, and Enterprise ChatGPT users. A Pro version is also rolling out to Pro, Business, and Enterprise tiers. OpenAI has now released multiple major models in the span of a few months, with no signs of slowing down.

Google has committed to investing up to $40 billion in Anthropic, making it one of the largest single bets on any AI company to date. The deal is structured in two parts; $10 billion goes in immediately at a $350 billion valuation for Anthropic, with the remaining $30 billion to follow if the company hits certain performance targets. This comes on top of a separate $5 billion investment Anthropic received from Amazon just this week, along with a broader agreement that will see Anthropic spend up to $100 billion on Amazon's cloud infrastructure over time.

A significant portion of Google's commitment comes in the form of compute, meaning access to Google Cloud's infrastructure and its tensor processing units, the specialised AI chips that are considered the strongest alternative to Nvidia's GPUs. Alongside the investment, Google Cloud is providing Anthropic with 5 gigawatts of computing capacity over the next five years, with room to expand further. For a company that has been facing user complaints about Claude performance limits in recent weeks, that kind of infrastructure commitment is as valuable as the cash itself.

Google has been backing Anthropic for a while, but the scale of this new commitment signals something bigger. Even as Google builds its own competing AI products through Gemini, it is simultaneously deepening its position as one of Anthropic's most important infrastructure providers. Earlier this month, Anthropic also announced a separate partnership with Google and chipmaker Broadcom to access additional gigawatts of TPU-based computing capacity from 2027 onward.

All of this is happening as Anthropic's valuation has been climbing fast. It stood at $350 billion as recently as February, but investors have since been offering to back the company at $800 billion or more. Reports also suggest Anthropic is considering an IPO as soon as October.

Your AI research assistant for discovering and analyzing scientific literature