iFlytek Spark X2.6 cloud-device integrated large model released, achieving breakthroughs in coding and agent capabilities based on fully domestic computing power
iFlytek released the iFlytek Spark X2.6 cloud-device integrated large model at the 1024 Developer Festival, claiming breakthroughs in coding and agent capabilities based on fully domestic computing power for the first time. The model adopts a 293B-A30B parameter MoE architecture, expands context length to 1M, and supports custom security policies by task scenario. The simultaneously released Spark X2.6-4B open-source on-device large model improves coding, agent, long-text and multilingual capabilities, adds three tiers of reasoning modes, and offers free reasoning for a limited time; the X2.6 series is now available for trial on the iFlytek Spark official website.
Why read this
AI score 68
iFlytek Spark X2.6 uses a 293B-A30B MoE architecture to expand context to 1M, and emphasizes that training is entirely based on domestic computing power, offering a reference for the actual progress of domestic chip training routes.
Topics
Body · AI translation
iFlytek Spark X2.6 Device-Cloud Integrated Large Model Released, First to Achieve Breakthroughs in Code and Agent Capabilities Based on Fully Domestic Computing Power
2026-10-22 14:40 IT Home - Wang Miao
IT Home, October 22 news: At this afternoon's iFlytek Global 1024 Developer Festival event, the iFlytek Spark X2.6 device-cloud integrated large model was officially released.
iFlytek Spark X2.6 claims to be the first to achieve breakthroughs in code and agent capabilities based on fully domestic computing power. iFlytek Chairman Liu Qingfeng said that Spark is the only one trained on fully domestic computing power, while all others are trained on Nvidia GPUs, and iFlytek's model is the smallest in size.
In addition, Liu Qingfeng said that iFlytek Spark X2.6 is the first to break through the training efficiency bottleneck of domestic chips for long-horizon tasks, with a dynamic parallel training technology for long-horizon tasks adapted to domestic platforms, solving the problem of low hybrid training efficiency caused by the memory bandwidth limitations of domestic chips.
Regarding the inference efficiency of domestic computing power clusters on long-horizon tasks, iFlytek has also achieved two important innovations:
One is distributed inference technology based on domestic heterogeneous computing power, solving the problem of slow training caused by the memory bandwidth limitations of domestic chips.
The other is the proposal of development technology based on domestic devices, where many things do not need to be recalculated and can directly use the original results.
In addition, the Spark X2.6-4B open-source on-device large model was officially released, the first to achieve a leap in on-device large model agent capabilities, with significant improvements in all benchmark scores compared with the X2.5-4B model.
IT Home learned from the launch event that since the Spark X2.5-4B on-device model was open-sourced on September 1, its download volume has entered the top four on-device models globally.
The iFlytek Spark X2.6 series models are now available for trial on the iFlytek Spark official website, debuting the Spark large model device-cloud integrated collaborative architecture.
iFlytek Spark X2.6 was trained based on domestic computing power, adopts a 293B-A30B parameter MoE architecture, with context length extended to 1M, and its code and agent capabilities reach the industry's top level. It also provides professional security protection and supports custom security policies by task scenario. The inference pricing is:
The iFlytek Spark X2.6-4B open-source on-device large model has achieved substantial improvements in code, agent, long-text and multilingual capabilities, added three tiers of inference modes, and its inference efficiency under the self-developed inference framework is improved compared with ollama. Inference pricing is free for a limited time.