AntAI releases Ling-3.2-flash: native multimodal model with 560B total parameters and 25B per-step activation.
AntAI unveils the native multimodal model Ling-3.2-flash, boasting 560B total parameters and 25B per-step activation, with support for 1M tokens of native long context. The model emphasizes visual feedback self-checking and adaptive recovery capabilities for anomalies.
Return to original text
A short introduction is retained here; the full text from the source website is not reposted. For specific wording and subsequent updates, please refer to the original article.
Open AIHOT original