​

DeepSeek V4 Pro and Flash API pricing: 30% off at peak and off-peak times

Rock API offers deepseek-v4-pro and deepseek-v4-flash at 70% of DeepSeek’s official price. The discount follows DeepSeek’s time-based billing: you pay the discounted peak rate during peak hours and the discounted off-peak rate during all other hours. Last updated: August 19, 2026

​

Rock API prices compared with official prices

The following prices are in USD per 1 million tokens. They apply the formula official rate × 0.7 to DeepSeek’s published rates effective August 16, 2026. DeepSeek defines peak hours as 01:00–04:00 and 06:00–10:00 UTC. All other hours are off-peak. That corresponds to 09:00–12:00 and 14:00–18:00 in Taipei (UTC+8). Check daylight-saving changes when converting the schedule for another region.

​

Why time-based 0.7x pricing matters

The 30% discount applies in both periods, while off-peak scheduling provides a second cost lever. For the same token category, Rock API’s off-peak price is 35% of DeepSeek’s official peak price because 0.5 × 0.7 = 0.35. For example, 100 million V4 Flash output tokens cost $132 at the official peak rate,$92.40 through Rock API at peak, or $46.20 through Rock API off-peak. Moving flexible work from peak to off-peak saves another 50% relative to the Rock API peak rate. Suitable workloads include:

• nightly evaluation and regression suites;
• repository indexing and codebase analysis;
• document extraction, summarization, and embedding preparation;
• queued agent jobs without a strict interactive deadline;
• synthetic-data and offline content pipelines.

Keep latency-sensitive user requests in their required time window. Schedule only workloads that can tolerate a queue, and monitor retries so a delayed batch does not spill into a peak period unexpectedly.

​

DeepSeek V4 Flash or V4 Pro?

Both models support a 1M-token context window, thinking and non-thinking modes, JSON output, tool calls, the Responses API, the Anthropic API, prefix completion, and non-thinking FIM completion. DeepSeek lists a maximum output of 384K tokens. DeepSeek currently identifies the served versions as DeepSeek-V4-Flash-0731 and DeepSeek-V4-Pro-0813. Its V4 release notes also document three thinking-effort levels: low, high, and max.

​

A practical cost-control plan

1. Route routine, high-volume requests to V4 Flash first.
2. Escalate only difficult or failed cases to V4 Pro.
3. Reuse stable prompt prefixes to improve cache-hit volume.
4. Queue flexible jobs for off-peak hours in UTC.
5. Record input, cached input, output, model, and execution window separately.
6. Compare cost per completed task, not only price per token.

​

Start using DeepSeek V4

1. Create a Rock API account.
2. Confirm that deepseek-v4-pro and deepseek-v4-flash are currently listed and review their live rates.
3. Create an API key and follow the authentication guide.
4. Send a small request using the quickstart, then verify the usage record and time band.
5. Read pricing and billing before moving scheduled production jobs.

​

Frequently asked questions

​

Is the 30% discount available during peak hours?

Yes. Rock API charges 70% of the applicable official rate in both peak and off-peak periods.

​

Is Rock API off-peak pricing 70% below the official peak price?

No. It is 65% below the official peak price: the official off-peak rate is half the peak rate, then Rock API applies 0.7x. The result is 35% of the official peak price.

​

Which timezone controls billing?

DeepSeek publishes the schedule in UTC. Peak hours are 01:00–04:00 and 06:00–10:00 UTC; every other hour is off-peak.