DeepSeek V4 Pro API Pricing
¥9.00 input · ¥27.00 output per 1M tokens at peak — 50% off off-peak. Structured reasoning with o1-level thinking.
DeepSeek V4 Pro is the flagship of the DeepSeek line: structured reasoning, complex code generation, and long-context mastery (1M tokens) with a Thinking Mode that produces chain-of-thought reasoning comparable to o1-level models — at a price point that makes it viable for high-volume production workloads.
Price per 1M tokens
DeepSeek V4 Pro · CNY official list price
| Tier | Rate |
|---|---|
| Input | Peak ¥9.00 · Off-peak ¥4.50 |
| Output | Peak ¥27.00 · Off-peak ¥13.50 |
| Cache hit | Peak ¥0.30 · Off-peak ¥0.15 |
DeepSeek lists official prices in CNY; USD uses ≈ 6.9 exchange rate
Best for
o1-level thinking
Thinking Mode produces chain-of-thought reasoning comparable to OpenAI o1-class models.
1M token context
Long-context tasks, whole-repo coding, and deep agent workflows fit comfortably.
~3.5× Flash output
About 3.5× V4 Flash on output (4.5× on input) at every hour — the extra quality is for the hard 10% of requests.
Complex reasoning pipelines, whole-repo code generation, and production workloads where structured, verifiable output offsets the ~3.5× price delta over V4 Flash.
Peak / off-peak billing
Peak (Beijing 09:00–12:00, 14:00–18:00) bills at full price; all other hours are 50% off. Cached input is ¥0.30 peak / ¥0.15 off-peak — 1/30 of the uncached rate.
V4 Pro vs the competition
Against GPT-5.6 Sol ($5/$30) and Claude Opus 5 ($5/$25), V4 Pro at ¥9.00/$27.00 peak competes on output quality for a fraction of the cost — and its ¥4.50 off-peak output undercuts every flagship on the market.
Frequently Asked Questions
What is the price of DeepSeek V4 Pro per million tokens?
DeepSeek V4 Pro bills ¥9.00 per million input tokens and ¥27.00 per million output tokens at peak (Beijing 09:00–12:00, 14:00–18:00); off-peak is 50% off at ¥4.50 / ¥13.50. Cache-hit input is ¥0.30 peak / ¥0.15 off-peak, again about 1/30 of the uncached rate.
When should I use V4 Pro instead of V4 Flash?
V4 Pro targets structured reasoning, complex code generation, and long-context tasks where output quality offsets the 3× price delta. The real question is whether Pro cuts retries, human review, or downstream errors enough to justify the gap for your workload.
Does V4 Pro support thinking mode and does it cost extra?
V4 Pro's Thinking Mode produces chain-of-thought reasoning comparable to o1-level models. Reasoning tokens are billed at the output rate, so factor them into your estimates. Routing batch jobs and cron workloads to Beijing evening/night windows cuts the bill further thanks to off-peak pricing.
Related model pricing
Compare this model against the rest of the lineup.
Estimate your exact cost
Model your own workload — input/output tokens, cache hit rate, and peak-hour share — against any model mix with the interactive pricing calculator.
Open the Pricing CalculatorRecommended Tools We ARE USING
We are using these tools ourselves for Development / Deployment. Check out for more details.
Opencode Go
You will receive $5 when subscribed, which directly offsets the 1st month's Go fee.
Could be the cheapest option for calling DeepSeek V4 Flash model in the world.
Vultr
RReceive $300 to test out VVultr platform
Referred user must be active 30+ days and spend $10–$25
Railway
Deploy anything without the complexity
Connect your repo, Railway handles the rest.
腾讯云 Tencent Cloud
Cloud servers, cloud databases, COS, CDN, SMS and other cloud products are on special offer now.
AWS high-quality alternative, we use COS as a replacement for S3
硅基流动 SiliconFlow
¥16 universal voucher for all platforms, valid for 180 days from the date of receipt
A comprehensive product matrix supporting the full-process implementation of AI applications.
Warp
One of the best AI terminal tools in the world
Interestingly, we can directly use various top-tier models in Warp without leaving the terminal.