DeepSeek makes V4-Pro's 75% API discount permanent, undercutting rivals 8x

DeepSeek has officially made permanent the 75% promotional discount for its V4-Pro API, locking in rates that heavily undercut Western competitors. According to an update on DeepSeek’s pricing page, the discount — originally tied to the V4 Preview launch on April 24, 2026 — will now become the standard price after the promo period ends on May 31, 2026 at 15:59 UTC. That means input tokens will remain at $0.435 per million and output tokens at $0.87 per million, instead of reverting to the list prices of $1.74 and $3.48 respectively.

To put this in perspective, DeepSeek V4-Pro is now roughly eight to nine times cheaper than OpenAI’s GPT-5.5 and Anthropic’s Claude Opus 4.7 for output tokens, while matching or exceeding them on specs: a 1 million token context window and up to 384,000 output tokens per request. Even their lightweight model, V4-Flash, costs just $0.14 per million input tokens — up to 100x cheaper than GPT-5.5 for inputs. This aggressive pricing strategy mirrors DeepSeek’s approach with V3 in late 2024, where a temporary discount eventually became the new standard. The move is already sparking debate on Hacker News and Reddit, with many noting it reshapes expectations for LLM inference costs and intensifies the price war between AI API providers.