DeepSeek Locks V4-Pro Discount at $0.87 per Million Tokens; Xiaomi Cuts MiMo-V2.5 by Up to 99% on May 26

According to official announcements, DeepSeek made its 75% V4-Pro API discount permanent on May 22, with output pricing locked at $0.87 per million tokens. Xiaomi followed on May 26 by slashing MiMo-V2.5 cached input costs by up to 99%, bringing the Pro model to $0.0036 per million tokens for cache hits—cheaper per token than most SMS character rates.

Meanwhile, American labs moved in the opposite direction. OpenAI doubled GPT-5.5's output price to $30 per million tokens at launch, while Anthropic's Claude Opus 4.7 shipped with a new tokenizer that can inflate actual costs by up to 35% despite unchanged rate cards. The pricing gap between DeepSeek V4-Pro and GPT-5.5 Pro, both scoring ~80% on SWE-Verified coding benchmarks, stands at 34x on output tokens. Chinese frontier models now trade at 15–30x lower cost than American counterparts for comparable performance.