DeepSeek: the right model and the right tool

Official prices, provider spreads, real cost math. Not a news site — a decision tool.

Models tracked
454
Providers tracked
45+
Widest price spread
18x
Off-peak discount
50%

What do you need?

Model pricing
Flash vs V4 Pro, peak vs off-peak, cache hits
Provider comparison
45 providers, up to 18x price spread
Cost calculator
Estimate your monthly spend in seconds
Prompt library
40 prompts, each with its token cost
Self-hosting
VRAM calculator and build-vs-buy break-even
Price history
Every DeepSeek price change since 2024

The two things that decide your bill

1. When you call

DeepSeek charges double during peak hours: UTC 01:00–04:00 and 06:00–10:00, Monday to Friday. In Beijing time that is 09:00–12:00 and 14:00–18:00 — Chinese working hours. Everything else, including weekends and public holidays, is half price.

Move your batch jobs to the evening and your cost drops by half, with no code change.

2. Who you call

DeepSeek publishes open weights, so dozens of providers resell the same model at wildly different prices. For deepseek-v4.1-flash-1017 we track 45 providers ranging from $0.025 to $0.450 per million input tokens — an 18x spread.

Cheapest is not always best. The lowest prices usually come with FP4 quantization, which hurts output quality. Our provider pages show quantization level next to price so you can pick deliberately.

Latest price data

ModelOff-peak inputOff-peak outputPeak multiplier
DeepSeek Flash$0.15$0.602.0x
DeepSeek V4 Pro$0.66$1.982.0x

USD per million tokens. Source: DeepSeek official API docs.

What this site is

We maintain price, provider and compatibility data for the DeepSeek ecosystem, and turn it into calculators and reference pages. The data updates weekly from public sources; every page shows when it was last verified.

We are not affiliated with DeepSeek. We do not resell API access. We do not offer consulting.