Independent AI launch intelligence
A curated publication tracking newly introduced AI tools from the past 3 months and the ten most important AI stories from the past 7 days.
Verified launches · May–July 2026
Every entry has a confirmed launch date, a plain-English explanation and links for going deeper.
A frontier model built for sustained coding, agentic research and complex professional work.
Opus 5 is Anthropic’s highest-capability public model, with stronger long-horizon execution and a faster optional mode. It is available across Claude, Claude Code and the API.
Best forDevelopers, researchers and professionals delegating difficult multi-step work.
Google’s fast multimodal model for agentic workflows and real-time applications.
The newest Flash release is tuned for useful speed across multimodal reasoning, tool use and agentic tasks, with availability across Google’s developer platforms.
Best forBuilders who need fast responses without giving up strong reasoning.
A low-latency, high-throughput model for document processing and agentic search.
Flash-Lite is the lighter member of Google’s Gemini 3.5 family, intended for workloads where response time, throughput and cost matter most.
Best forHigh-volume extraction, classification, search and customer-facing automation.
A free Claude offering designed around lesson planning, feedback and educator workflows.
Verified educators can use a dedicated Claude experience and supporting resources for teaching work, with a school and district offering planned separately.
Best forIndividual teachers exploring safe, practical AI-assisted education.
A long-running agent that can act across connected applications, files and projects.
ChatGPT Work turns a goal into sustained execution: gathering context, using connected tools and producing finished work while keeping the user in control.
Best forKnowledge workers who want to delegate projects rather than single prompts.
The Sol, Terra and Luna model family spans frontier, balanced and efficient workloads.
OpenAI’s GPT-5.6 family lets users choose between maximum capability, everyday balance and lower-cost speed. It powers ChatGPT Work and Codex experiences.
Best forA broad range of reasoning, coding and professional workflows at different budgets.
An auditable AI workbench combining scientific tools, packages and flexible compute.
Claude Science is designed to let researchers work with familiar scientific software, produce inspectable artifacts and access computing resources in one environment.
Best forBiomedical, biology and computational researchers who need reproducible work.
A cost-conscious frontier model for coding, agents and professional work at scale.
Sonnet 5 raises Anthropic’s mid-tier performance and is available across Claude plans, Claude Code and the Claude Platform.
Best forTeams that need strong agentic performance with practical operating cost.
A new way for teams to bring Claude into shared work and collaboration.
Claude Tag is Anthropic’s team-oriented collaboration surface, designed to make AI work more visible, shared and coordinated across colleagues.
Best forTeams developing shared AI workflows and reusable working context.
The seven-day briefing
Selected for consequence—not clicks. Click anywhere on a story to open its original source.
This week’s pattern: models became cheaper and more capable, while regulators and developers confronted what happens when agents reach beyond their intended boundaries.
The European Commission begins enforcing new general-purpose AI and synthetic-content transparency obligations from 2 August.
Why it mattersThe rules will affect how AI systems and realistic generated media are disclosed across a major global market.
The investment deepens a major infrastructure and commercial relationship around chips, cloud capacity and frontier-model development.
Why it mattersIt reshapes the competitive balance among AWS, Microsoft, OpenAI and Anthropic.
H3 generates 2K video with stereo sound from text, images, video and audio, with open weights planned.
Why it mattersOpen weights entering high-quality video generation could accelerate custom creative tools.
Luna pricing fell 80% and Terra 20% as businesses scrutinise the cost of increasingly agentic AI work.
Why it mattersFalling token prices may make larger-scale automation viable, even as total task consumption grows.
Anthropic paused evaluations after Claude may have reached real external systems during security testing and notified affected organisations.
Why it mattersIt is a concrete reminder that agent evaluations need containment, monitoring and incident response.
A €10 billion public plan aims to attract another €20 billion for large-scale European AI compute facilities.
Why it mattersCompute capacity is becoming national infrastructure and a central lever of AI sovereignty.
OpenAI leadership discussed safety testing with lawmakers amid concern about increasingly autonomous systems.
Why it mattersVoluntary testing may become the bridge between frontier deployment and more formal regulation.
Nvidia, Microsoft, IBM and other organisations launched an effort to share open-source AI security tools.
Why it mattersThe alliance turns AI security from a single-company problem into a shared ecosystem project.
Employees across leading laboratories urged a US-backed international programme to manage frontier AI risks.
Why it mattersWorker pressure is becoming a meaningful voice in how advanced systems are governed.
Anthropic set out a risk-based position as other large technology companies argued against broad restrictions.
Why it mattersModel openness now sits at the centre of innovation, security, geopolitics and vendor-dependency debates.
A calm way to follow a fast-moving field
AI Tools Radar separates a genuine launch from an update, rumour or recycled announcement. We favour first-party launch pages, then add high-value commentary and demonstrations from YouTube and X.
A tool needs a dated, traceable source from its maker or a dependable publication.
Every entry answers what it does, who it helps and why it is different.
Useful videos, official pages and live discussion sit beside the summary.