New GPT-5.6 Tiers, Claude Sonnet 5, and Grok 4.5 Land Amidst White House AI Access Directives
Major AI labs pushed new model tiers and agent capabilities this week, intensifying the race for specialized, production-ready AI skills.
New here? The 60-second primer
Skills are packaged expertise you install into an AI assistant: Claude Code has skills and plugins (from Anthropic's directory and open git marketplaces), ChatGPT has custom GPTs (the GPT Store), and Google's Gemini has Gems (shareable custom experts). Every rank below comes from captured data — official install counts, GitHub star growth, and the GPT Store's own trending list — not from opinion.
The week
This week saw significant movement in the AI skills landscape with several major model releases and a notable shift in government oversight. OpenAI's GPT-5.6 family, comprising Sol, Terra, and Luna tiers, became generally available on July 9, offering a 1.05M-token context window for diverse workloads.
Anthropic launched Claude Sonnet 5, its agentic mid-tier model, with introductory API pricing, and debuted Claude Science for researchers on July 12. Anthropic also released Claude for Teachers on July 14, offering free premium access to K-12 educators. xAI shipped its Grok 4.5 model and open-sourced Grok Build, enabling developers to create and deploy AI agents.
Google began rolling out its Gemini Spark personal AI agent to Google AI Ultra subscribers, though the expected Gemini 3.5 Pro release slipped. Google also announced a transition from Gemini CLI to Antigravity CLI, impacting consumer users.
A major development reported on July 17 was the White House dictating early access to frontier AI models from OpenAI, Anthropic, and Google, a power previously held by the labs. This follows Anthropic's Fable 5 global access restoration on July 1, after a Commerce Department suspension.
Top Claude plugins — official installs
| Rank | Plugin | Installs | Signal |
|---|---|---|---|
| 1 | Frontend Design | 1,081,079 | baseline week |
| 2 | Superpowers | 965,863 | baseline week |
| 3 | Code Review | 422,373 | baseline week |
| 4 | Context7 | 406,106 | baseline week |
| 5 | Skill Creator | 367,694 | baseline week |
| 6 | Code Simplifier | 336,733 | baseline week |
| 7 | Github | 309,067 | baseline week |
| 8 | Playwright | 307,688 | baseline week |
| 9 | Claude Md Management | 276,947 | baseline week |
| 10 | Feature Dev | 249,751 | baseline week |
Rising Claude skill repos — GitHub momentum
| Repo | This Week | Total Stars |
|---|---|---|
| obra/superpowers | +2,871 stars | 257,986 |
| zarazhangrui/frontend-slides | +336 stars | 25,978 |
| K-Dense-AI/claude-scientific-skills | +337 stars | 31,281 |
| anthropics/claude-cookbooks | +310 stars | 49,200 |
| ykdojo/claude-code-tips | +83 stars | 9,364 |
ChatGPT GPT Store — official trending
| Rank | Gpt | Maker |
|---|---|---|
| 1 | Scholar GPT | awesomegpts.ai |
| 2 | Fitness, Workout & Diet - PhD Coach | Newgen PhD |
| 3 | image generator | naif alotaibi |
| 4 | Consensus | consensus.app |
| 5 | Write For Me | puzzle.today |
| 6 | Code Copilot | promptspellsmith.com |
| 7 | Video AI by invideo | invideo.io |
| 8 | Logo Creator | — |
| 9 | Canva | — |
| 10 | Photo Realistic Image GPT | genigpt.net |
Gemini Gems worth knowing
| Gem | What It Does | Source |
|---|---|---|
| RayStock AI | Full-spectrum Taiwan stock analysis | awesome-gemini-gems |
| RayBuffett | Berkshire Hathaway 13F tracker | awesome-gemini-gems |
| RayMarketing | Social media post writer | awesome-gemini-gems |
| RayReply | Difficult-customer response advisor | awesome-gemini-gems |
| RayTranslator | Context-aware multilingual translation | awesome-gemini-gems |
| RayLabor | Taiwan labor law expert | awesome-gemini-gems |
| RayAntiScam | Fraud detection assistant | awesome-gemini-gems |
| RayClassical | Classical Chinese translator | awesome-gemini-gems |
| RaySketch | Hand-drawn sticker studio | awesome-gemini-gems |
| RayDream | Dream interpretation | awesome-gemini-gems |
Pattern watch
A clear pattern emerging this week is the acceleration towards specialized, multi-agent systems and the orchestration of AI agents. The concept of "loop engineering," described by Boris Cherny (Claude Code's creator) and Peter Steinberger (OpenClaw's creator), gained traction as a key method for AI agents to iterate on software development.
This trend aligns with Anthropic's 2026 report, which highlights the shift from writing code to orchestrating agents that write code, and the move from single-agent workflows to coordinated teams of specialized agents. The Agensi marketplace, which crossed 2,000 skills and 50,000 monthly visitors, further underscores the rising demand for ready-made, specialist AI capabilities.
Try this week
To explore the growing ecosystem of AI skills, consider trying a Claude plugin. You can install a plugin by opening the 'Customize' menu in the left sidebar of Claude, navigating to the 'Plugins' tab, clicking 'Browse plugins,' and then selecting 'Install' on your chosen option. This allows Claude to integrate with external tools and services, expanding its capabilities for specific tasks.
Reporting + analyst voices: grounded via Google Search at publish time.