The new native AI model from Ant Ling consists of 560B total parameters and 25B active parameters to support medical intelligence and complex visual understanding. Designed for long-horizon agents and 3D interaction, Ling-3.2-flash outscored Opus 5 (max) in release evaluations on the ZEROBench(Sub), Vision2Web, and OmniDocBench1.5 benchmarks.
The model reached a rank of #10 on the DesignArena Image to Frontend Arena with an Elo score of 1238, placing it alongside systems from Meta, DeepSeek, and Anthropic. Ant Ling priced the model’s input at $0.30 and output at $0.90 per 1M tokens, a rate the DesignArena leaderboard noted is the lowest for any model providing this level of performance.
