Dateline: April 17, 2026 | Next update: April 24, 2026
A landmark week for Anthropic: Claude Opus 4.7 launched as the new flagship model, Cowork went generally available on Mac and Windows, and Claude Code shipped major quality-of-life updates including session recaps and push notifications.
Claude / Anthropic
Claude Opus 4.7 — new flagship model
Opus 4.7 is now the most capable generally available Claude model, with major gains in agentic coding, vision, instruction-following, and long-running task reliability. It replaces Opus 4.6 at the same price point.
Opus 4.7 is generally available. Benchmarks: 64.3% on SWE-bench Pro, 87.6% on SWE-bench Verified, 69.4% on Terminal-Bench 2.0, and 64.4% on Finance Agent v1.1. Stronger at catching its own logical faults during planning, self-verifying outputs, and handling underspecified requests. Also available in Amazon Bedrock on AWS-managed infrastructure with zero operator access. Note: Opus 4.7 includes API breaking changes vs Opus 4.6 — review the migration guide before upgrading.
Model string: claude-opus-4-7 | Pricing: $5/$25 per MTok (unchanged) | 1M token context | Breaking API changes vs Opus 4.6 — see Migrating to Claude Opus 4.7 | Also available on Amazon Bedrock (us-east-1) via Bedrock-Mantle endpoints
Best for: Complex coding, long-horizon agentic tasks, financial analysis, vision workloads
Model deprecations — action required
Three models are now on deprecation timelines. Haiku 3 retires this Saturday. Sonnet 4 and Opus 4 (original) are newly deprecated with a June retirement.
Claude Haiku 3 (claude-3-haiku-20240307) retires April 19, 2026 — two days away. Migrate to Haiku 4.5 now. Claude Sonnet 4 (claude-sonnet-4-20250514) and Claude Opus 4 (claude-opus-4-20250514) are newly deprecated, retiring June 15, 2026. Migrate to Sonnet 4.6 and Opus 4.7 respectively.
Haiku 3 retirement: April 19, 2026 | Sonnet 4 + Opus 4 retirement: June 15, 2026 | 1M context beta for Sonnet 4.5 and Sonnet 4 also retires April 30, 2026
Best for: API developers — immediate action required for Haiku 3 users
Claude Cowork — generally available
Cowork has left preview and is now generally available. Enterprise teams get new analytics, monitoring, and access control tools. The mobile app now supports fully interactive connectors for live charts and visual assets in-chat.
Cowork is now GA on macOS and Windows via Claude Desktop. New enterprise capabilities: usage analytics, OpenTelemetry support for monitoring Cowork activity, and role-based access controls with SCIM group sync from your identity provider. The Claude mobile app can now render fully interactive apps — live charts, diagrams, and shareable assets — directly in conversation.
Analytics via Enterprise Analytics API | OpenTelemetry: monitor Cowork activity with your existing observability stack | RBAC: manually assign roles or sync groups via SCIM | Interactive mobile connectors: live charts, sketchpads, shareable assets
Best for: Enterprise teams managing agent workflows, mobile users needing rich visual output
ant CLI — new command-line API client
Anthropic launched the ant CLI this week — a dedicated command-line client for the Claude API. It integrates natively with Claude Code and lets you version API resources as YAML files, making API work scriptable and repeatable.
ant CLI is now available — faster API interaction from the terminal, native Claude Code integration, and YAML-based versioning of API resources including Managed Agents endpoints.
Covers Claude API + Managed Agents (beta header: managed-agents-2026-04-01) | YAML versioning for API resources | See CLI reference in Claude Docs
Best for: Developers and DevOps teams working directly with the Claude API
Claude Code — session recap, push notifications, Advisor tool
Claude Code had a busy week with several point releases covering new features, reliability fixes, and Windows improvements.
Session recap feature added — provides context when returning to a long session, invocable with /recap or auto-triggered on return. Push notification tool added — Claude can send mobile push notifications via Remote Control. 1-hour prompt cache TTL opt-in added (ENABLE_PROMPT_CACHING_1H). New /tui fullscreen command for flicker-free rendering. Advisor tool launched in public beta — pair a fast executor model with a high-intelligence advisor model for long-horizon agentic tasks at lower cost.
New env vars: ENABLE_PROMPT_CACHING_1H, FORCE_PROMPT_CACHING_5M | /recap command | /tui fullscreen | /focus command (replaces Ctrl+O for focus view) | Advisor tool beta header: advisor-tool-2026-03-01 | Windows: CLAUDE_ENV_FILE + drive-letter path fixes | Extended-thinking indicator now shows rotating progress hint
Best for: Developers, CI pipelines, long-running agentic coding sessions
Plans and Pricing
No changes to consumer or API pricing this week. Opus 4.7 launches at the same $5/$25 per MTok price as Opus 4.6.
Opus 4.7: $5/$25 per MTok | Sonnet 4.6: $3/$15 per MTok | Haiku 4.5: low-cost tier | Managed Agents: contact enterprise sales
ChatGPT / OpenAI
Dateline: April 10–17, 2026 | Next update: April 24, 2026
Over the past week, OpenAI has focused on improving multimodal consistency, system responsiveness, and enterprise integrations.
GPT-5.3 Standard — default model
Improved response consistency and faster outputs.
Reduced latency across standard queries and better instruction adherence.
Context ~128k | Output ~4k–8k | Minor inference optimisations
Best for: General use
GPT-5.3 Pro — high-reasoning model
More stable long-form reasoning and fewer dropped steps in complex prompts.
Improved coherence in extended analytical tasks.
Context ~200k (est.) | Slight speed improvements | More stable chain-of-thought execution
Best for: Deep analysis
GPT-5.3 Mini — fallback model
Faster responses for lightweight tasks.
Smarter routing between Mini and Standard models.
Context ~64k | Improved routing logic
Best for: Quick tasks
Agent Mode
Handles multi-step workflows.
Better task persistence across longer workflows and reduced need for re-prompting.
Improved session memory handling | More reliable task chaining
Best for: Task delegation
Deep Research
Combines browsing and reasoning.
Improved synthesis of multiple sources into single coherent outputs.
Better ranking + summarisation pipeline | Slight latency increase
Best for: Research
Memory & Projects
Persistent context across chats.
Improved prioritisation of recent and task-relevant memories.
Refined memory weighting system | Reduced irrelevant recall
Best for: Ongoing workflows
Advanced Voice Mode
Improved conversational flow.
More natural interruption handling and smoother transitions between speakers.
Latency reduction in turn-taking | Better voice activity detection
Best for: Voice interaction
Other Features
Minor improvements to Canvas stability and responsiveness. Early-stage improvements to multimodal (image + text) alignment. No major updates to Sora, Codex, or pricing.
What this means
This week's updates are not about flashy new releases, but about making the system more reliable, faster, and more usable in real-world workflows. Three things matter:
First, latency and consistency improvements (especially in GPT-5.3 Standard) suggest OpenAI is prioritising everyday usability. The model is becoming less "impressive but unpredictable" and more "quietly dependable," which is critical for professional use.
Second, Agent Mode and memory improvements point to a clear strategic direction: turning ChatGPT into a tool that can handle ongoing, multi-step work with less supervision. This is a shift from answering questions to actually doing tasks.
Third, better routing between Mini and Standard models shows continued optimisation on the cost-performance tradeoff. This signals that OpenAI is refining how intelligence is allocated behind the scenes — important for scaling.
Overall, this is a consolidation week: fewer headline features, more infrastructure-level improvements. These are the kinds of changes that typically precede larger feature rollouts.
Gemini
Date: April 16, 2026 | Next update: April 23, 2026
Following the major Gemma 4 launch last week, Google DeepMind has shifted focus to specialized multimodal expansions, including the debut of Gemini 3.1 Flash TTS and significant updates to Robotics and Workspace Personalization.
Gemini 3.1 Flash TTS — new preview
A cost-efficient, expressive, and "steerable" text-to-speech model now available for developer preview in the Gemini API. (Launched April 15, 2026)
Best for: Voice & audio apps
Gemini Robotics-ER 1.6 — new preview
Released April 14, 2026, this specialized model features improved spatial reasoning and "instrument reading" capabilities for complex physical environments. Note: the 1.5-preview version is scheduled for shutdown on April 30, 2026.
Best for: Robotics & hardware development
Gemma 4 open models
The gemma-4-26b and gemma-4-31b models, released on April 2, remain the primary focus for developers building lightweight, open-weight applications.
Gemini "Personal Intelligence" — beta rollout
A major beta rollout began on April 16, 2026 for AI Pro and Ultra subscribers in the U.S. Gemini can now connect directly to Gmail, Photos, YouTube, and Search to provide proactive, personalized help based on your specific context.
Best for: Personal productivity
Step-by-step Guided Learning
A new "Guided Learning" mode is now available to all users (via the "Learn" chip). It moves beyond simple Q&A to break down complex subjects step-by-step for deeper comprehension.
Best for: Education & training
Google Vids upgrade
As of April 2, 2026, users can create and edit AI-powered videos at no cost, featuring the new Nano Banana 2 image generation integration.
Chrome integration
Launched on April 14, 2026 — users can now turn their most frequent AI prompts into one-click tools directly within the Chrome browser.
Plans and migration deadlines
Whisk to Flow final countdown: With only two weeks remaining, users are urged to migrate all assets from Whisk to Flow before the April 30, 2026 deadline. Media not moved by this date will be permanently deleted.
Model retirement: The gemini-2.5-flash-lite-preview-09-2025 model was officially shut down this month; all developers should now be using gemini-3.1-flash-lite-preview.
Billing & usage: Google introduced revamped Usage Tiers and project-level spend caps in AI Studio to help developers better manage costs for high-volume 3.1 Flash workloads.
Microsoft Copilot
Dateline: April 17, 2026 | Next update: April 24, 2026
Microsoft 365 Copilot — Scatter image effect
A Scatter image effect has been added to Copilot's image editor in Chat and Create modules. This effect applies a modern, stylized look to AI-generated or uploaded images, making it easier for users to enhance visuals quickly.
Scatter image effect added to Copilot Chat + Create editor.
Roadmap ID: 537288 | Platforms: Windows, Web, Android, iOS, Mac | Integrated into Copilot Chat + Create editor
Best for: Users needing fast, stylized image enhancements across devices
Role-based Copilot offerings — Release Wave 1 (2026)
Microsoft published the April 2026 release wave plan for role-based Copilot offerings. This includes new features tailored to specific roles across Dynamics 365 and Microsoft 365, with updates rolling out progressively through September.
Release wave 1 plan published — covers Dynamics 365 and Copilot role-based features. Release Planner and downloadable PDF available. Updated continuously throughout the release cycle.
Best for: Enterprise teams planning adoption of role-based Copilot features
Windows 11 — April 2026 Patch Tuesday update
Microsoft rolled out the April Patch Tuesday update for Windows 11, bringing accessibility improvements, display and hardware enhancements, and File Explorer quality-of-life updates.
Applies to Windows 11 versions 25H2 and 24H2. Two-phase rollout strategy with Controlled Feature Rollout (CFR) — features delivered immediately for some devices, gradually for others.
Best for: Windows 11 users seeking improved accessibility and smoother system performance
Plans and Pricing
No changes to Copilot consumer or enterprise pricing this week. Microsoft 365 Copilot pricing remains unchanged. Role-based Copilot features included in enterprise subscriptions.
This week's Copilot update emphasizes new creative tools (Scatter effect), enterprise role-based expansions, and system-level improvements in Windows 11.