AI Capability Releases: What's New in June 2026

Cheaver's monthly guide to the latest updates from Claude, ChatGPT, Copilot, and Gemini

For June, every major platform shipped significant updates: from new frontier models and agentic frameworks to personal finance tools, background automation agents, and deep enterprise integrations.

• Anthropic released Claude Sonnet 5 and lifted export restrictions on its most powerful models.

• OpenAI introduced GPT-5.6 in limited preview and made Codex Remote available to all.

• Microsoft's Copilot Cowork went globally available, positioning it as a true autonomous work engine.

• And Google's Gemini lineup expanded with Gemini 3.5 Flash, the Omni multimodal model, and a 24/7 background agent called Spark. If you've been watching the Google AI development, June delivered on the promises made at Google I/O and beyond.

Claude (Anthropic)

Claude Sonnet 5: The Most Agentic Sonnet Yet

Anthropic released Claude Sonnet 5, described as its most agentic Sonnet model to date. Sonnet 5 brings stronger reasoning, tool use, coding, and autonomous task handling to the table. It now serves as the default model in Claude Code and ships with a native 1M-token context window, providing enough to ingest entire codebases or long-form document repositories in a single session. Promotional pricing is set at $2/$10 per million tokens through August 31, making it a compelling option for teams building agentic pipelines.

Export Controls Lifted: Fable 5 and Mythos 5 Go Global

As of June 30, Anthropic lifted export controls on its two most powerful models — Claude Fable 5 and Claude Mythos 5. Starting July 1, Fable 5 became available globally on Claude Platform, Claude.ai, Claude Code, and Claude Cowork. This represents a significant expansion of access to frontier-level capabilities for international users and enterprises.

Claude Apps Gateway for Bedrock and Google Cloud

Enterprises running Claude on Amazon Bedrock or Google Cloud gained a powerful new self-hosted control plane via the Claude Apps Gateway. It provides corporate SSO, centrally enforced policy, role-based access controls, per-user cost tracking, and spend caps. This is the enterprise governance layer many organizations have been waiting for!

Trusted Devices for Remote Control Admins

Team and Enterprise plan administrators can now require members to verify their device before initiating or viewing remote Claude Code sessions. This Trusted Devices feature adds a hardware-level verification step to remote development workflows, reducing the risk of unauthorized access to sensitive codebases.

MCP Tunnels, Live Settings, and Self-Hosted Sandboxes

The Claude Developer Platform received several Research Preview features: MCP (Model Context Protocol) tunnels, self-hosted sandboxes for Claude Managed Agents, and the ability to update MCP server and tool settings during active sessions without restarting. Large tool outputs are now automatically spilled to a sandbox file rather than bloating the context window.

Claude in Microsoft Foundry: Generally Available

Claude's integration with Microsoft Foundry reached general availability, giving Azure-native developers direct access to Claude models within the Foundry toolchain.

Apple Foundation Models Support

Claude now interoperates with Apple's Foundation Models framework, available for iOS 27, iPadOS 27, macOS 27, visionOS 27, and watchOS 27. Developers can pass typed outputs from Apple's on-device features directly into Claude requests, enabling tighter on-device / cloud hybrid AI workflows.

ChatGPT (OpenAI)

GPT-5.6 Limited Preview: Sol, Terra, and Luna

OpenAI began a limited preview of GPT-5.6, introducing three specialized reasoning modes codenamed Sol, Terra, and Luna. The new models target stronger performance in reasoning, coding, biology, and cybersecurity. This preview marks the next step in OpenAI's model roadmap following the consolidation of earlier GPT-4.5 and GPT-5.2 variants into GPT-5.5.

Model Consolidation: GPT-4.5 and GPT-5.2 Retired

As part of its ongoing model consolidation, OpenAI retired GPT-4.5 from ChatGPT on June 26, with all existing conversations automatically migrating to GPT-5.5. GPT-5.2 models followed suit on June 12. Users won't notice a disruption, as conversations will simply continue on the newer, more capable GPT-5.5, but this signals OpenAI's intent to streamline its model lineup.

Codex Remote: Generally Available on All Plans

Codex Remote (OpenAI's tool for running agentic coding tasks on a connected local machine) is now generally available across all ChatGPT plans. Users can start or continue work on a connected Mac or Windows host, review progress, and approve actions directly from their phone. A new DigitalOcean Droplet Workspace plugin lets Codex automatically provision a cloud VM and connect it as a remote workspace, making serverless agentic coding accessible without a local machine.

Personal Finance Experience Rolls Out to Plus Users

ChatGPT's personal finance experience (which helps users track spending, review budgets, and understand financial trends) is rolling out to Plus users in the U.S. on web and iOS, and is now available on Android for Pro and Plus users. This positions ChatGPT as a conversational personal finance assistant, competing with dedicated finance apps.

Active Sessions Security Feature

A new Active Sessions panel lets users review all sessions associated with their ChatGPT account and sign out of any they don't recognize. The panel surfaces details such as device type, app, approximate location, sign-in time, and trusted-device status, thereby bringing consumer-grade session management in line with what enterprise security teams expect.

Memory Management Improvements

Users now have finer control over ChatGPT's memory. The updated memory summary page allows deletion of individual memories, and the option to select 'Delete and turn off memory' gives users a clean-slate reset from the three-dot menu.

Dictation Gets 10%+ More Accurate

OpenAI upgraded ChatGPT's voice dictation model, delivering at least a 10% reduction in word error rate for top languages in noisy environments like public spaces and open-plan offices. The improvement is particularly meaningful for users who rely on voice input as their primary interaction method.

Copilot (Microsoft)

Copilot Cowork: Generally Available Worldwide

Microsoft's most significant June announcement was the global general availability of Copilot Cowork, which is their agentic system designed to plan, execute, and deliver completed work rather than drafts or recommendations. Users define the task; Cowork handles everything from start to finish and returns a finished deliverable. Cowork runs on usage-based billing, and administrators can manage costs through a new Cost Management Dashboard in the Microsoft 365 admin center, with budgets and per-user spend caps available.

Claude Now Available in Copilot Chat

Anthropic's Claude is now a selectable model option within Microsoft 365 Copilot Chat. Users can choose Claude for complex analysis, document understanding, and structured content generation, which just provides users with another third-party model option alongside Microsoft's own offerings. This reflects the emerging multi-model strategy that enterprises are increasingly demanding.

Chat / Cowork Toggle

The Microsoft 365 Copilot app now features a simple toggle between Chat mode and Cowork mode, giving users a clear on-ramp to agentic work without having to navigate menus. This UX decision reflects Microsoft's intent to make agentic workflows a mainstream choice rather than a power-user feature.

Intelligent Model Selection in Cowork

Cowork can now automatically select the model best suited to the task at hand. OpenAI's GPT-5.5 Thinking is used for deep research with citations, Anthropic's models handle visual work like PowerPoint and graphics, and other configurations are available depending on the task type. This model-agnostic routing is a significant architectural step for enterprise AI workflows.

New Plugins: Enosix, Harvey, Miro, monday.com and More

Cowork's plugin ecosystem expanded in June with new integrations including Enosix (SAP connectivity), Harvey (legal AI), LSEG (financial data), Miro (visual collaboration), and monday.com (project management). Additionally, Cowork now sends push notifications for long-running tasks, so users don't need to monitor the app while work is in progress.

Power BI Integration via Work IQ

Microsoft 365 Copilot can now reason over Power BI enterprise data and return grounded, cited answers from Power BI reports and semantic models, all through natural language. Users can ask business questions without building queries or knowing SQL, and Copilot will surface the relevant data directly from existing reports.

Gemini (Google)

Gemini 3.5 Flash: Frontier Performance for Agents

Google released Gemini 3.5 Flash as part of its latest model family, combining frontier intelligence with action-oriented capabilities. The model delivers strong performance for agents, coding, and complex long-horizon tasks. Gemini 3.5 Pro is currently in testing with a release expected next month, positioning Google to maintain competitive parity with OpenAI and Anthropic at the top end of the capability spectrum.

Gemini Omni: Any Input, Any Output

Gemini Omni is a new multimodal model capable of generating any output from any input, starting with video. Gemini Omni is rolling out in the Gemini app to all Google AI subscribers globally (18+). The model marks Google's most serious push into native video generation integrated directly into its AI assistant platform.

Native Visual Models: Image Generation from Video

Two new native visual models reached general availability: Gemini 3.1 Flash Image and Gemini 3 Pro Image. The Flash Image model adds video-to-image generation support, so users can pass video files or YouTube URLs alongside text prompts to generate thumbnails, movie posters, or custom infographics. This opens entirely new creative workflows for content creators.

Gemini 3 Deep Think and Deep Research

Gemini 3 Deep Think began rolling out to Google AI Ultra subscribers, offering a dedicated reasoning mode for math, science, logic, and hard multi-step problems. Deep Research positions Gemini as an autonomous research assistant that breaks a prompt into sub-questions, scans multiple sources, and assembles multi-page reports.

Gemini Spark: Your 24/7 Background Agent

Gemini Spark is a 24/7 personal AI agent that continues working even when your phone is locked. As a cloud-based agent, Spark can execute custom user-defined workflows independently. The feature is also coming to the Gemini macOS app, extending Spark's reach to desktop automation across files and applications.

Daily Brief: Personalized Morning Digest

The Daily Brief feature pulls together a user's inbox, calendar, and priority tasks into a clear morning overview. It goes beyond aggregation by suggesting next steps and prioritizing action items, essentially functioning as your chief-of-staff that sets the user's agenda for the day.

New Integrations: Canva, Dropbox, Instacart, OpenTable, Zillow Rentals

Google expanded Gemini's integration ecosystem with five new consumer and business services: Canva (design), Dropbox (file storage), Instacart (grocery delivery), OpenTable (restaurant reservations), and Zillow Rentals (apartment search). These additions make Gemini significantly more useful as a daily life assistant beyond productivity tasks.

Gemini in Chrome on Android

Gemini is launching directly in Chrome on Android in late June, initially available on devices with 4GB of RAM or more with their language set to English-US. This browser-native integration gives users access to Gemini without switching apps, making AI assistance available at the point of browsing.

What Does This Mean for Users?

1. Agentic AI Is No Longer Experimental — It's the Default

Between Claude Sonnet 5's expanded agentic capabilities, Codex Remote reaching all ChatGPT plans, Copilot Cowork's global launch, and Gemini Spark's background execution, June made one thing crystal clear: AI that acts on your behalf, not just answers your questions, is now a standard feature, not a beta perk. If you haven't started experimenting with delegating multi-step work to an AI agent, your peers likely have.

2. Your AI Can Now Work While You Sleep

Gemini Spark and Copilot Cowork can both operate as background agents, completing tasks when your device is locked or when you step away. This means you can assign a research task, a document generation job, or a data analysis request before a meeting and return to a finished deliverable instead of just a draft waiting for your attention.

3. Multi-Model Workspaces Are Becoming the Norm

Microsoft's Copilot now lets users choose Claude or OpenAI models depending on the task. Google's Gemini integrates with Canva, Dropbox, and business tools. Claude works within Microsoft Foundry and Apple's ecosystem. You no longer need to choose one AI platform because the best setups will mix models and tools to match the task at hand.

4. AI Is Moving Into Your Financial Life

ChatGPT's personal finance rollout to Plus users represents a meaningful step toward AI-driven personal financial management. Meanwhile, Copilot's Power BI integration brings natural-language business intelligence to the enterprise. Whether you're a solo entrepreneur or part of a large organization, AI is becoming a realistic option for understanding your numbers without a data analyst.

5. Video and Visual Creation Are Getting Dramatically Easier

Google's Gemini Omni and video-to-image generation capabilities mean that content creators can now generate thumbnails, movie posters, and marketing visuals directly from video input. No coding or prompting voodoo required. Generate amazing results, just using natural language. Combined with Canva integration, these tools reduce the barrier to professional-quality visual content significantly.

6. Enterprise Security and Governance Are Catching Up to AI Capability

This month saw meaningful progress in AI governance: Anthropic's Claude Apps Gateway delivers corporate SSO and spend caps; Trusted Devices adds hardware verification to remote development; OpenAI's Active Sessions panel surfaces session transparency; and Copilot's Cost Management Dashboard gives IT administrators budget controls. AI is maturing from a shadow IT tool into a governable enterprise system.

7. Dictation and Voice Are Getting Significantly More Reliable

OpenAI's 10%+ improvement in dictation accuracy in noisy environments, combined with Gemini's browser-native presence in Chrome on Android, points to a near future where voice-first AI interaction is viable in everyday settings, not just a quiet home office.

The Bottom Line

June 2026's theme was delivery. The big announcements from earlier in the year, including agentic frameworks, background execution, multi-model flexibility, and enterprise governance, moved from preview to production. For business professionals and power users, the practical implication is simple: the tools to delegate complex, multi-step work to AI are now live, stable, and available at mainstream price points. The competitive gap between platforms is narrowing, which means the real differentiation will increasingly come from how well you integrate these capabilities into your actual workflows. The organizations that win in the second half of 2026 will be the ones who stopped evaluating and started operationalizing.

This monthly roundup goes out every month, covering exactly what changed and what it means for your work. If you found this useful, forward it to a colleague who's still on the fence about AI, or leave a comment below with the update that caught your attention most. And if you're not yet subscribed, this is your reminder: the AI landscape shifts fast, and a monthly digest is the most efficient way to stay current without spending hours in the weeds.