Anthropic has an unreleased model that infers users’ goals from sparse instructions

Anthropic reportedly has developed a new AI model that can infer users’ goals from minimal instructions, potentially surpassing current models in reasoning capabilities. This information emerged from a social media post, suggesting internal advancements at the lab responsible for the Claude AI series. While this model has not been publicly named or benchmarked, it indicates Anthropic’s continuous efforts to compete at the frontier of AI technology. Anthropic’s current lineup, including Claude Opus 5.5, holds significant positions on AI leaderboards, and the news of this unreleased model may affect its standing in the ongoing competition with Google and other industry leaders.

Source: Crypto Briefing