DeepSeek V3.1 API price hike and off-peak discounts removed
From September 6, DeepSeek-V3.1 input cache-miss ¥4, output ¥12 per 1M tokens, cache-hit ¥0.5, off-peak discounts removed.
On September 6, DeepSeek officially enforced the new V3.1 API pricing. After adjustment, input is ¥0.5 per 1M tokens on cache hit and ¥4 on cache miss; output is ¥12 per 1M tokens. Compared with the previous V3 (¥2 input miss, ¥8 output), the overall price rose noticeably.
At the same time, DeepSeek cancelled its off-peak discount. Previously users enjoyed a 50% discount from 0:30 AM to 8:30 AM daily; that policy ended on September 6.
Despite the increase, DeepSeek-V3.1 remains far cheaper than closed-source flagships like GPT-5 (about ¥8.9 input, ¥71 output per 1M tokens) and Claude Opus 4.1 (about ¥107 input, ¥536 output), retaining a cost advantage. Multiple institutions noted that in 2025 domestic LLM discounts expired one after another, with the industry shifting from pure price competition to value competition.