DeepSeek API Pricing & Token Economics
DeepSeek has revolutionized AI token economics with high-capacity Mixture-of-Experts (MoE) architectures (DeepSeek-V4.1-Flash and DeepSeek-V3). DeepSeek features an automated diurnal pricing schedule that grants an unconditional 50% discount during off-peak compute hours (Mon-Fri 00:00–01:00, 04:00–06:00, 10:00–24:00 UTC and all weekends), alongside an up to 98% prefix cache discount ($0.006/1M peak, $0.003/1M off-peak on V4.1 Flash; $0.014/1M on DeepSeek-V3).
Automatic prefix caching discounts hits by up to 98% ($0.006/1M peak, $0.003/1M off-peak on V4.1 Flash; $0.014/1M on DeepSeek-V3).
Standard API execution rules. Refer to individual endpoint rate limits for asynchronous processing.
50% discount automatically applied during off-peak windows: Mon-Fri 00:00–01:00, 04:00–06:00, 10:00–24:00 UTC and 24 hours on weekends.
Active DeepSeek Models(3 models verified)
All 39 Models →Calculate your monthly DeepSeek bill
Simulate prompt caching hit rates, batch discounts, and output token costs with zero server logging.