deepseek decision makes it the cheapest among the smart models

deepseek has announced a steep price cut for its v4-pro model api. once its 75%-off promo ends on may 31, 2026, v4-pro settles at a quarter of list price – roughly $0.44 per million input tokens and $0.88 output (converted from yuan)

• the output price is the headline: at $0.88, v4-pro undercuts every other major chinese model by a wide margin, which matters since output usually drives real api bills

• rivals stay close on input. minimax m2.7 actually beats deepseek there ($0.279) but charges more for output ($1.20). http://z.ai's glm 5.1 and xiaomi's mimo-v2.5-pro sit mid-tier (around $1 in, $3 out), moonshot's kimi-k2.6 lands at $0.95 in / $4.00 out, while alibaba's qwen 3.7 max is the premium pick at $2.50 in / $7.50 out.

another shot in china's llm price war