{T}
TokenMath.net
Official Provider HubHangzhou, China

DeepSeek API Pricing & Token Economics

DeepSeek has revolutionized AI token economics with high-capacity Mixture-of-Experts (MoE) architectures (DeepSeek V4 and V4.1 Flash). DeepSeek features an automated diurnal pricing schedule that grants an unconditional 50% discount during off-peak compute hours (16:30–08:30 UTC), alongside a 90% prefix cache discount.

Prompt Caching Policy

Automatic prefix caching discounts hits by up to 90% ($0.014/1M on V4.1 Flash).

Batch / Asynchronous Rules

Standard API execution rules. Refer to individual endpoint rate limits for asynchronous processing.

Special Billing Conditions

50% discount automatically applied between 16:30 and 08:30 UTC every day.

Active DeepSeek Models(2 models verified)

All 30 Models →
ModelContextInput / 1MCached / 1MOutput / 1MBatch DiscountActions
DeepSeek-V4-Flash
deepseek-v4-flash
1.05M$0.44$0.014$1.3250% offSpecs →
DeepSeek-V3
deepseek-chat
128k$0.14$0.014$0.28N/ASpecs →

Calculate your monthly DeepSeek bill

Simulate prompt caching hit rates, batch discounts, and output token costs with zero server logging.

Open Cost Calculator →