The latest model releases, pricing changes, and product news that actually affect which tools you should use — updated as things change.
Every fact is linked to a primary source.

The Claude maker's S-1 reveals 1,300% revenue growth, an $80.7B net loss, and a cofounder-controlled share structure.

Runway's new tool generates on-brand ads, publishes to Meta, Google and TikTok, and iterates on what earns spend.

Eligible ChatGPT Pro and Business subscribers can claim up to $500 a month in free API credits to build apps and agents.

Google DeepMind's new world model simulates photorealistic scenes you steer with natural language, with native audio and dialogue.

Describe the tool you want in plain language and Runway builds a custom image, video, and audio app around your brand kits and workflows.

Anthropic's new agent comes to you with a morning brief, works in the background, and checks before it acts on your behalf.

Zapier folds its standalone Agents product into a single AI step — with per-tool approvals, task-based pricing, and full run history.

Maia-guided improvements, fallback AI connections, Claude Sonnet 5.5 and GPT-6.1 Sol, SSO routing, and MCP audit tools.

A vision-language-action model, an embodied-reasoning model, and an on-device variant for humanoids and multi-robot work.

Faster responses, stronger tool use, and multi-agent execution for Pro users — plus "dots" that triage incidents overnight.

Google opens its frontier model to developers and enterprises — a unified 1M-token window, long decode, and $2/$10 intro API pricing.

Anthropic bundles monthly API credits into Max and Team plans — $100–$200 for Max, up to $500 pooled for teams.

Around 75% cheaper to run than Haiku 4.5, with input from $0.10 per million tokens and a first-ever adjustable effort setting.

Google's new frontier model brings a 1M-token output limit and defensive-cybersecurity skills, rolling out first to trusted cyber defenders.

Nearly matches GPT-6 Astra on coding, computer use, and professional work — at one-fifth of Astra's token prices.

After an agent bypassed DNS filtering in a training sandbox, OpenAI halted training, evaluation, and tool-use inference.

30%+ faster than Sonnet 5 and up to 30% cheaper per task, with a 70.6% Terminal-Bench score vs Sonnet 5's 10.3%.

Korean ships to everyone, plus live style previews, default parameters, and a Create feed redesign.

Anthropic's new life sciences lab reports Claude found "ART" — a new enzyme system with CRISPR-like repeats.

Credits now work across the web app and Runway Dev, with a single invoice and centralized usage analytics.

Matches Fable 5.1 at 40% lower cost — $4/$20 API pricing and a 1M-token context window.

Design custom voices from prompts, 100+ languages, and replication from 30 seconds of audio.

Two new GPT-6 models with API prices 50% lower than GPT-5.6 — live in ChatGPT Work and Codex now.

Sharper aesthetics, smarter personalization, and a new Edit Model that replaces Omni Reference.

Two new live dialogue models — one for fluid chat, one for multi-step reasoning — with 97-language support.

Anthropic merges Cowork into one Claude and launches Docs and Slides in beta on paid plans.

Draft, summarize, revise, and format documents from a sidebar inside Word — on all plans, including Free.

Shared projects, a pooled credit balance, and Agent Connectors for Figma, Dropbox, and Notion.

Anthropic's beta plugin brings accounts, pipeline, and 37 sales skills into Claude.

More capable and ~25% cheaper than Fable 5, with big gains in coding and research.

1M-token context, stronger coding, and a security-focused Flash Cyber — at Flash pricing.

The Codex harness for any developer, plus full-duplex voice in the API at $0.05/min.

What the rollout means for ChatGPT plans, and the computer-use and coding gains to expect.

A 230M-URL legal index, firm-grade privacy controls, and 26 partner plugins.