GPT-5.4 Mini Rolls Out to All Free and Go Users via the Thinking Feature
If you are on OpenAI's free tier or Go plan, you now have access to GPT-5.4 mini through the Thinking option in the composer menu. This is a meaningful step up for users who have been limited to standard models: the mini variant brings reasoning capabilities that were previously gated behind paid plans, making AI-assisted problem solving more accessible to everyday users.
For Plus, Pro, and Business subscribers, the model also functions as an automatic fallback when GPT-5.4 Thinking hits rate limits during high-demand periods, helping to maintain uninterrupted access. The rollout effectively extends reasoning-tier performance to a broader audience without requiring a subscription upgrade.
ChatGPT Retires Its Legacy Deep Research Mode
OpenAI has switched off the legacy version of its Deep Research mode as of 26 March 2026. If you have been using the older interface, your historical conversations and results remain accessible, but any new research sessions will now run on the current Deep Research experience. For most users this will be a seamless transition, since the updated mode has been the default for some time.
It is worth noting this if you have been relying on the legacy version for specific workflows, as any behavioural differences between the two versions will now be permanent.
Long Pastes in ChatGPT Now Auto-Convert to Attachments for Paid Users
ChatGPT will now automatically convert any pasted text longer than 5,000 characters into an attachment rather than inserting it directly into the message composer. This lands for Plus, Pro, and Business subscribers first. For anyone regularly dropping in long documents, code blocks, or transcripts, this removes a common friction point: the composer stays clean and readable, and the content is treated as a properly scoped context file rather than an unformatted wall of text.
It is a small quality-of-life change but one that makes a noticeable difference in day-to-day use.
Google Drive Becomes a Single Unified Connector in ChatGPT
OpenAI has consolidated its Google file integrations inside ChatGPT under one Google Drive app, replacing the previous arrangement of separate connectors for Docs, Sheets, and Slides. New users can now connect a single Google Drive integration to access all three formats, removing the need to authenticate separately for each product. For existing users who had already set up individual apps, the unified experience will improve consistency.
If you build or use workflows that involve pulling Google Workspace files into ChatGPT sessions, this simplification is worth noting.
Claude Mobile App Gains Live Interactive Apps with Charts, Diagrams, and Shareable Assets
Anthropic has updated the Claude iOS and Android apps to support fully interactive embedded applications within the chat interface. This means that when Claude generates a chart, diagram, or data visualisation, you can now interact with it directly in the mobile app rather than receiving a static image. You can also build and share lightweight assets without leaving the conversation.
For anyone who uses Claude on mobile for analytical or creative work, this closes a meaningful gap with the desktop experience and brings the app closer to a fully capable workbench rather than a text-first interface.
Claude Code Ships Major Windows Update with Native PowerShell Support in Opt-In Preview
Claude Code has received a substantial update focused on Windows developers, marking a significant step towards first-class Windows support. The release includes a PowerShell tool available as an opt-in preview, which enables Claude Code to run PowerShell commands natively on Windows rather than routing through WSL. Beyond the new tool, the update brings faster startup times, smarter MCP and workspace handling, fixes to clipboard behaviour on Windows and WSL, voice mode on Windows, and a fix for extra VS Code windows opening at startup.
If you have been using Claude Code primarily on macOS or Linux, this release signals that Anthropic is actively investing in making the Windows experience production-grade.
Anthropic identified a low-rate error issue affecting requests to Claude Opus 4.6 and Sonnet 4.6 on 27 March 2026. The company flagged the problem via its status page. Source: status.anthropic.com
Industry Themes
Today's updates reinforce two patterns that have been building across the last few weeks. First, the major labs are competing on access breadth as much as raw capability: OpenAI extending reasoning-tier models to free users and Anthropic bringing interactive apps to mobile both reduce the gap between entry-level and power-user experiences. If you have been recommending AI tools to non-technical users, the free-tier story is noticeably stronger today than it was a month ago.
Second, developer tooling on non-Mac platforms is catching up quickly. Anthropic's Claude Code Windows release with native PowerShell support reflects a broader industry recognition that a significant share of professional developers work on Windows, and that feature parity matters for adoption. Expect more Windows-first announcements from other labs in the coming weeks.
Finally, the absence of new releases from Google DeepMind, xAI, Meta, and Mistral today highlights that even in a fast-moving market, product cadence is uneven. The next major model releases from Grok 5 and others remain on the horizon but have not landed yet.