AI inference cost war intensifies

Xiaomi permanently slashed MiMo-V2.5 API prices up to 99% (cache hits), following DeepSeek's permanent 75% cut on V4 Pro (34x cheaper vs GPT-5.5). This signals rapidly falling AI inference costs, which could improve Creatify's margins or enable more aggressive pricing against rivals like JoggAI, while also pressuring Western model providers Creatify relies on. The trend accelerates commoditization of video generation models, forcing differentiation on features rather than price. Latest: DeepSeek flash pricing tweet reinforces the cost war trend.