Search
Baidu Ling-2.6-flash Officially Released: Achieving SOTA-Level Agent Performance with 104B Parameters and Only 1/10 Token Consumption
5 hour ago /
Read about 0 minute
Author：小编
On April 22, 2026, Ant Group's Baidu large model officially released Ling-2.6-flash, an Instruct model with a total of 104B parameters and 7.4B activated parameters, focusing on 'Token efficiency.' The model achieved the best performance in its size category across multiple Agent-related benchmark tests. In the Artificial Analysis evaluation, it completed tasks using only 15M tokens, approximately 1/10 of the consumption of other models. The API pricing for Ling-2.6-flash is set at $0.1 per million tokens for input and $0.3 for output, with availability now open and a one-week free trial offered. Previously, its anonymous test version, 'Elephant Alpha,' reached a daily invocation volume of 100B on the OpenRouter platform.
-
C114 Communication Network -
Communication Home
7 X 24 Track global technological trends
