DeepSeek’s API lists two models: deepseek-flash (served by DeepSeek-V4.1-Flash-1017) and deepseek-v4-pro (DeepSeek-V4-Pro-0813). Flash is far cheaper and the only one with vision, while both share a 1M-token context and a 384K maximum output. Peak hours cost double, so check the pricing page before you build on these numbers.
The short answer
- Flash output: $0.6 per million tokens off-peak and $1.2 at peak.
- Pro output: $1.98 per million tokens off-peak and $3.96 at peak.
- Peak pricing: off-peak rates are half the peak rates.
- Vision: only deepseek-flash supports it.
- Shared features: JSON output, tool calls, the Responses API and the Anthropic API format, with non-thinking and thinking modes (thinking is the default).
How do the two models compare?
| Per 1M tokens | deepseek-flash | deepseek-v4-pro |
|---|---|---|
| Model version | DeepSeek-V4.1-Flash-1017 | DeepSeek-V4-Pro-0813 |
| Context / max output | 1M / 384K | 1M / 384K |
| Input, cache hit (off-peak / peak) | $0.003 / $0.006 | $0.022 / $0.044 |
| Input, cache miss (off-peak / peak) | $0.15 / $0.3 | $0.66 / $1.32 |
| Output (off-peak / peak) | $0.6 / $1.2 | $1.98 / $3.96 |
| Vision | Yes | Not supported |
| Concurrency limit | 2,500 | 500 |
By our arithmetic from the table, Pro output costs 3.3 times Flash output ($1.98 against $0.6), and Pro cache-miss input costs 4.4 times Flash ($0.66 against $0.15). Flash also allows five times the concurrency (2,500 against 500). Pick Pro only if your own tests show Flash fails your task.
When are peak and off-peak hours?
According to DeepSeek’s pricing page, peak hours are 01:00 to 04:00 and 06:00 to 10:00 UTC, Monday to Friday, excluding Chinese public holidays. All other times are off-peak, at half the peak rate. The page gives no end date for this scheme.
Our arithmetic: one million Flash output tokens cost $0.6 off-peak and $1.2 at peak, so shifting a batch job out of peak hours saves $0.6 per million output tokens.
How do you connect?
Apply for an API key on the DeepSeek Platform. The OpenAI-format base URL is https://api.deepseek.com. The Anthropic-format base URL is https://api.deepseek.com/anthropic.
So you can point the OpenAI or Anthropic SDK, or compatible software, at DeepSeek by changing the configuration.
How are Search and Fetch billed?
Search and Fetch are $0.75 per 1,000 successful requests each. Calls made through DeepSeek Harness, Claude Code or Codex are currently free. Per-request billing starts at 00:00 Beijing time on 1 November 2026.
Which model names are retired?
The names deepseek-v4-flash and deepseek-v4-flash-vision-exp are retired. Requests under those names are still accepted, served by DeepSeek-V4.1-Flash-1017 and billed at the Flash price.
What to do next
- Open the Models & Pricing page and confirm the current prices.
- Price peak and off-peak hours separately.
- If you still use the retired model names, switch to deepseek-flash.
- Start with Flash, and move to Pro only for tasks where Flash fails your own tests. This is our suggestion.
Questions and answers about the DeepSeek API
Is the DeepSeek API OpenAI-compatible?
Yes. The OpenAI-format base URL is https://api.deepseek.com. It also offers an Anthropic-format endpoint at https://api.deepseek.com/anthropic.
Does Flash support images?
Yes. The page lists vision as supported for deepseek-flash and not supported for deepseek-v4-pro.
How much does Flash output cost?
$0.6 per million tokens off-peak and $1.2 at peak.
Can I still call deepseek-v4-flash?
Yes, for now. Requests under the retired name are served by DeepSeek-V4.1-Flash-1017 and billed at the Flash price.
When are Search and Fetch billed?
From 00:00 Beijing time on 1 November 2026, at $0.75 per 1,000 successful requests each.
Terms in plain language
- Token: a small piece of text that a model reads or writes.
- Context: how much text the model can take in at once.
- Cache hit: input the service has already processed, which costs less.
- Base URL: the web address your code uses to reach the API.
Sources
Facts read on 22 October 2026.
Part of the guide: AI Model Pricing and API Cost Guides
