iFlytek releases Spark X2.6 end-cloud unified model at $0.09 CNY-equivalent per million input tokens
At its 1024 Developer Day, iFlytek released Spark X2.6, a 293B-total/30B-active mixture-of-experts model with 1M token context, priced at 0.9 CNY per million input tokens, 0.08 cached and 3.6 output, with a 4B distilled version open-sourced.
iFlytek released the Spark X2.6 model on October 22 at its 1024 Developer Day. It uses a mixture-of-experts architecture with 293 billion total parameters and 30 billion active, supporting a context window of up to 1 million tokens.
Pricing is 0.9 CNY per million input tokens, 0.08 CNY per million cached input tokens and 3.6 CNY per million output tokens. On cached reads, real cost on long-context workloads is materially lower than input-billed pricing.
On efficiency, the company reports a 4x training efficiency gain and 4.5x inference efficiency gain, with KVCache hit rate rising from 63% to 95.7% and expert routing accuracy at 96.1%. Benchmarks include SWE Multilingual at 95.0 and Terminal Bench 2.1 at 83.2. A distilled Spark X2.6-4B version was open-sourced in parallel, trained entirely on domestically produced compute.