US Bans Anthropic's Fable 5 as Google, Nvidia Drop Powerful New Models

The last 24 hours saw a flurry of major AI model releases and a landmark government action. Google unveiled two image-capable models: Gemini 3 Pro Image (priced at $2.00 input / $12.00 output per 1M tokens, 66K context) and Gemini 3.1 Flash Image ($0.50 / $3.00, 131K context). Nvidia countered with Nemotron 3 Ultra 550B, a massive 550-billion parameter model available via Nebius at $1/$3 per 1M tokens with a 1M context window. Other notable releases include Z.ai's GLM-5.2, which is being hailed as the top frontend coding model, Qwen3.7 Plus from Alibaba (1M context, $0.32/$1.28), and MiniMax M3 (1M context, $0.30/$1.20). Cohere also launched North Mini Code, a free coding-focused model. These models push the boundaries of cost efficiency and capability, especially in image generation and long-context reasoning.

In a dramatic twist, the US government banned the release of Anthropic's Claude Fable 5, citing safety concerns. The ban came amid ongoing tension between Anthropic and regulators over the model's advanced capabilities, which some argue could be misused. Anthropic had opened Fable 5 to the public just days earlier with safegguards, but the government intervened. This marks one of the first explicit bans on a frontier AI model by a major government, setting a precedent for future regulation. The incident overshadows the broader release cycle, as developers and enterprises now face a landscape where powerful models are both accelerating and facing increasing scrutiny.