AI News
What is new in AI, all in one place: models, products, funding and policy.
Anthropic released Claude Sonnet 5 on June 30, 2026, with introductory API pricing of $2 per million input tokens and $10 per million output tokens through August 31, 2026, after which pricing moves to $3/$15.
On July 8, 2026, OpenAI launched two full-duplex voice models in ChatGPT: GPT-Live-1, the new default for paid users, and GPT-Live-1 mini, the default for free users. Both replace Advanced Voice Mode (source: OpenAI official announcement, 2026-07-08).
On July 7, 2026, GitHub made the GitHub Copilot desktop app available on every plan, including Copilot Free and GitHub Education, on macOS, Windows, and Linux.
Cursor released version 3.11 on July 10, 2026, adding side chats, conversation search backed by a local index, redesigned project and repo pickers, and three cloud agent hooks.
Runway launched Agent Skills for all plans on July 2, 2026: single commands that build ad campaigns, create commercials, and localize ads. The release lands one week after the Agent 2.0 upgrade on June 25.
On July 7, 2026, YouMind shipped an update: Skill homepages get custom cover images and local video demo reels, the editor adds lossless image upscaling, and Claude Sonnet 5 with adaptive thinking is now built in.
OpenAI, the maker of ChatGPT, released the GPT-5.6 model family on July 9, 2026, in three tiers—Sol, Terra, and Luna—with input pricing from $1 to $5 per million tokens.
The Council of the EU gave its final green light to the Digital Omnibus on AI on June 29, 2026, completing the first amendments to the EU AI Act and pushing back the high-risk compliance deadline that was set for August 2, 2026.
China's Interim Measures for the Administration of Anthropomorphic AI Interaction Services took effect on July 15, 2026. The rules were jointly issued in April by the Cyberspace Administration of China and four other departments, and they cover AI companion and emotional-interaction chatbots.
Moonshot AI released Kimi K3 on July 17, 2026: a natively multimodal sparse MoE model with 2.8 trillion total parameters and a 1M-token context window. The company says it will publish the full weights by July 27, 2026.