iFlytek Spark X2.5 Model Officially Released: 293B-A30B MoE, Trained on Fully Domestic Platform
iFlytek released the Spark X2.5 model, with a 293B-A30B MoE architecture, trained on a fully domestic platform, supporting 256K context and over 200 languages, already launched on the iFlytek Xingchen MaaS platform.
Key Takeaways
- Key Highlight：iFlytek released the Spark X2.5 model, with a 293B-A30B MoE architecture, trained on a fully domestic platform, supporting 256K context and over 200 languages, already launched on the iFlytek Xingchen MaaS platform.
- Innovation & Tech：Highlights advancements in Spark, X2.5, Model, demonstrating rapid progress in model capabilities.
- Industry Impact：Reported via IT之家 (CN), offering actionable signals for developers and technology leaders.
iFlytek officially released the Spark X2.5 multilingual text generation model, adopting a MoE architecture with 293B total parameters and 30B activated parameters, supporting a 256K context length. The model is trained on a fully domestic platform, covering over 200 languages, with significantly enhanced code generation and agent capabilities.
The release of Spark X2.5 reflects progress in domestic large models achieving autonomous and controllable training. Completing large-scale MoE model training on a fully domestic computing platform helps reduce dependence on imported chips and serves as a demonstration for the completeness and security of the domestic AI industry chain.
The model has been launched on the iFlytek Xingchen MaaS platform, with pricing at 1.6 RMB per million tokens for input, 0.24 RMB per million tokens for cache hit, and 6 RMB per million tokens for output. This price range is at a mid-level among similar large models and may encourage more enterprises to adopt domestic large models for application development.
With the improvement in multilingual and agent capabilities, Spark X2.5 is expected to expand its application scope in scenarios such as overseas services, multilingual customer service, and complex code generation. Its actual effectiveness and commercial performance in the future deserve continued attention.
This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.
Industry Insights & Analysis
As artificial intelligence rapidly evolves, breakthroughs surrounding Spark, X2.5, Model, Officially are shifting toward scalable, robust real-world implementations.
Driven by both open-source ecosystems and proprietary model architectures, the integration between compute optimization, data engineering, and agentic workflows is accelerating. This development provides a strategic benchmark for upcoming AI tooling and developer workflows.