iFlytek Spark X2-Flash Model Officially Released, Supporting 256K Long Context and Agent Framework Compatibility

iFlytek's Spark X2-Flash model was officially released today with open API access. It uses a MoE architecture, has 30B total parameters, supports up to 256K context length, and has significantly improved capabilities in agents and code. It was trained on the Huawei Ascend 910B cluster.

AstronClaw and others have already integrated the model. Tests show it performs close to trillion-parameter models in tasks like deep research reports and Skill management, with token consumption only one-third of similar large models, reducing developer costs. It is also compatible with mainstream Agent frameworks like OpenClaw, facilitating complex application development.