AI Round-up: July 1, 2026

Anthropic dominated the snapshot. Claude Sonnet 5 launched as a more agentic Sonnet model, GitHub and Perplexity quickly added support, and the same window brought a separate access story: US export controls on Fable 5 and Mythos 5 were lifted, with Anthropic saying Fable 5 would return globally with new cybersecurity classifiers.

The broader research-and-agent story stayed active around science and enterprise workflows. Anthropic introduced Claude Science and Reuters covered it as a scientific research platform, OpenAI published GeneBench-Pro for biological-data agents, Microsoft moved Service Agent to general availability and framed agent skills as trainable assets, while Google DeepMind and Runway pushed new image/video generation releases.