Question 1

What models are available?

Accepted Answer

DeepSeek offers two main models: deepseek-chat (DeepSeek-V3) and deepseek-reasoner (DeepSeek-R1).

Question 2

What are the context lengths?

Accepted Answer

Both models support up to 64K tokens of context.

Question 3

What is Chain of Thought (CoT)?

Accepted Answer

CoT is the reasoning content deepseek-reasoner provides before outputting the final answer. It helps the model to break down complex problems into smaller steps.

Question 4

What is Context Caching?

Accepted Answer

Context Caching can reduce costs by caching previously seen context, resulting in cheaper fees for cached inputs. Ideal for chatbots and long conversations.

Question 5

How do the discount periods work?

Accepted Answer

DeepSeek offers significant discounts (50% off for deepseek-chat, 75% off for deepseek-reasoner) during off-peak hours (16:30-00:30 UTC daily).

Question 6

How do deduction rules work?

Accepted Answer

The expense = number of tokens × price. The corresponding fees will be directly deducted from your topped-up balance or granted balance, with a preference for using the granted balance first when both balances are available.

DeepSeek API Discount Timer (Archived)