IT News on September 30th: Ant Birdling has released the Ling-3.1-flash model today, which boasts approximately 560B total parameters, with each Token activating around 25B parameters; the model's context window has a maximum of 1M, allowing it to accommodate longer documents, code, and task histories.
Officials stated that during the development process, continuous optimizations were made around tasks such as general intelligent agents, search, daily office work, and software development. Additionally, the model's capabilities were continuously enhanced in scenarios such as medical, financial, and materials science research.
In addition, Ling-3.1-flash will offer a two-week free trial; considering the comprehensive service costs, the duration of the model service during the free trial period is 256K hours.
After the free trial period ends and it transitions to a paid service, it is planned to open up 1 million contexts. At that time, the team also plans to make the code open-source and continue to improve model performance.
The IT family's model achieved the following scores in multiple task evaluations:
