Claude Fable 5 Returns With New Cybersecurity Classifiers

Anthropic redeploys its flagship model globally after U.S. government talks — coding tasks face restrictions, rate limits reset across all tiers, and platforms rush to integrate.

AnthropicAI · July 1, 2026

Anthropic announced the global redeployment of Claude Fable 5 following a series of productive conversations with the U.S. government. The model returns with a new set of classifiers designed to target and block more cybersecurity tasks. In the near term, some routine coding tasks will be restricted under the new safety framework. U.S. Commerce Secretary Howard Lutnick formally withdrew export controls on both Fable 5 and Mythos 5 in a letter to Anthropic, with conditions attached.

Pro, Max, and Team users can allocate up to 50% of their weekly usage quota toward Fable 5 at no extra cost through July 7, after which the model will be billed via usage credits. Enterprise Standard seats receive no free quota and use credits from day one. Cloud platform availability on AWS, Google Cloud, and Microsoft Foundry is expected in the coming weeks. The rate limits for all users have been reset, allowing developers to immediately resume building. Fable 5 is also available again in Cursor, where it leads all models on CursorBench but carries the highest per-task cost. Other platforms including v0 by Vercel and Perplexity Computer have similarly reintegrated the model.

Early access user Ethan Mollick described Fable 5 as far surpassing all previously available public models. In his testing, the model excelled at complex tasks: generating multi-page specifications, producing academic papers, composing 10-page rhyming poems, and autonomously launching sub-agents to retrieve and cross-validate over 2,200 flight schedules, railway timetables, and road data across multiple countries — sustaining hours of autonomous work. When feedback demanded improvements, the model deployed adversarial agent groups to acquire remote region transportation data, demonstrating a new level of autonomous cross-validation.

Zhipu AI Ships ZCode, the Official IDE for GLM-5.2

ZCode integrates AI agents with existing tools for planning, coding, review, and deployment. Available on macOS, Windows, and Linux with BYOK support for existing subscriptions and APIs. GLM Coding Plan subscribers receive a 1.5× usage quota within the IDE.

Runway Ships Nano Banana 2 Lite at Warp Speed

Runway launched Nano Banana 2 Lite, delivering high-quality images at blazing speed without sacrificing quality. The model is available in Runway Studio and can be invoked through Runway Agent. v0 by Vercel has also integrated NB2L for image generation directly in its prompt bar.