Search
iFLYTEK has unveiled its Spark X2.5 large model, which adopts a Mixture of Experts (MoE) architecture with a total of 293 billion parameters, though only 3 billion parameters are activated during inference. The model has achieved significant breakthroughs in code generation and agent interaction, while also enhancing general capabilities such as mathematical computation and semantic understanding. It supports a context length of 256K and has been deeply optimized for domestic computing environments. Additionally, Spark X2.5 introduces two lightweight edge-side models, 4B and 1.7B, which are open-source and capable of processing contexts up to 1 million tokens in length, enabling local deployment across various scenarios. Currently, the series of models is fully available on the iFLYTEK Open Platform.
-
C114 Communication Network -
Communication Home
