DeepSeek Pricing 2026: Free App and API Price List
What DeepSeek's free app includes, every API rate with peak and off peak prices, cache discounts and worked examples, from DeepSeek's API docs.
By , research by the BestBots team · Updated Oct 18, 2026
Fact checked against each company's published pricing, documentation and verified user ratings. How we score
DeepSeek's chat app is free, with no paid plan. Its API, as listed in DeepSeek's docs in October 2026, costs $0.30 per million input tokens and $1.20 per million output tokens for deepseek-flash at peak, and half that off peak. deepseek-v4-pro costs $1.32 and $3.96 at peak. Cache hits are up to 50 times cheaper.
- Starting price
- $0
- Pricing model
- free
- Free plan
- Yes
- Free trial
- None listed
- Free plan
- Yes, the chat app is free
- Paid from
- No subscription; API from $0.15 per 1M input tokens off peak
- Top model on paid plan
- No subscription; API tops out at DeepSeek-V4-Pro
- Context window
- 1M tokens (API)
- Image generation
- Not listed
What the free app includes
| Feature | Free app |
|---|---|
| Price | $0, no subscription |
| Reasoning | DeepThink |
| Web search | Yes |
| File uploads and vision | Yes |
| Image generation | Not listed |
| Team plan or admin controls | None |
The app has no paid tier, so there is no higher usage plan to buy. Reviewers on the App Store report chat sessions that expire on long projects, which is the practical limit of free use. Data from the app is stored in China and used for training unless you email an opt out, per DeepSeek's privacy policy.
API price list
| Model | Input, cache hit | Input, cache miss | Output |
|---|---|---|---|
| deepseek-flash, peak | $0.006 per 1M | $0.30 per 1M | $1.20 per 1M |
| deepseek-flash, off peak | $0.003 per 1M | $0.15 per 1M | $0.60 per 1M |
| deepseek-v4-pro, peak | $0.044 per 1M | $1.32 per 1M | $3.96 per 1M |
| deepseek-v4-pro, off peak | $0.022 per 1M | $0.66 per 1M | $1.98 per 1M |
Both models list a 1M token context window and up to 384K output tokens, with thinking mode, JSON output and tool calls. deepseek-flash adds vision; deepseek-v4-pro does not.
| Tool | Price |
|---|---|
| Search and Fetch | $0.75 per 1,000 successful requests (free through DeepSeek Harness, Claude Code or Codex until November 1, 2026) |
What the pricing terms mean
Peak and off peak. Peak hours are 01:00 to 04:00 and 06:00 to 10:00 UTC, Monday through Friday, excluding Chinese public holidays. That is the evening and early morning in the Americas and the working day in East Asia. Everything else, including all weekend hours, is off peak at half price. Batch jobs scheduled for US business hours or weekends run at the lower rate.
Cache hit versus cache miss. If the opening of your prompt (a system prompt, a long document, a codebase) matches a recent request, those tokens bill at the cache hit rate. On deepseek-flash at peak, that is $0.006 instead of $0.30 per million, a 50 times discount. Apps that reuse a long fixed prompt see most of their input billed at the cache rate.
Billing. You top up a balance, and usage draws from any granted balance first, then your topped up funds. There is no monthly fee.
Worked examples
| Workload | Model and timing | Monthly cost |
|---|---|---|
| 10M input, 2M output tokens | deepseek-flash, peak | $5.40 |
| 10M input, 2M output tokens | deepseek-flash, off peak | $2.70 |
| 10M input, 2M output tokens | deepseek-v4-pro, peak | $21.12 |
| 10M input, 2M output tokens | deepseek-v4-pro, off peak | $10.56 |
| 100M input (80% cache hits), 10M output | deepseek-flash, peak | $18.48 |
The last row is 80M cached tokens at $0.006 ($0.48), 20M uncached at $0.30 ($6) and 10M output at $1.20 ($12). Caching cuts the input bill from $30 to $6.48.
Hidden costs and fine print
- Peak timing. If your users are in East Asia or your jobs run at peak hours, you pay the full rate.
- Output tokens dominate. Output costs four times input on deepseek-flash and three times on v4-pro. Reasoning in thinking mode produces more output tokens.
- Search is extra after November 1, 2026 at $0.75 per 1,000 requests.
- Data location. The API, like the app, is operated by a company that stores data in China. Some employers restrict it.
How it compares
| Provider | Model | Input per 1M | Output per 1M |
|---|---|---|---|
| DeepSeek | deepseek-flash (peak) | $0.30 | $1.20 |
| DeepSeek | deepseek-v4-pro (peak) | $1.32 | $3.96 |
| xAI | grok-4.3 | $1.25 | $2.50 |
| xAI | grok-4.7 | $2 | $6 |
| Anthropic | Claude Sonnet 5.5 | $2 | $10 |
| Anthropic | Claude Opus 5.5 | $4 | $20 |
On published per token rates, deepseek-flash is the cheapest option among the labs we track, and off peak it halves again. For consumer comparisons, the free app competes with ChatGPT Free, Gemini Free and Meta AI rather than paid plans. The BestBots AI Price Index tracks seat plans; DeepSeek has none. Rival API and plan prices are in Claude pricing and Grok pricing.
Which option to pick
- Personal, non sensitive use: the free app.
- High volume, cost sensitive API work: deepseek-flash, scheduled off peak, with a fixed prompt prefix for cache hits.
- Harder reasoning tasks on the API: deepseek-v4-pro.
- Anything confidential or regulated: a provider with business data terms. See our DeepSeek review for the privacy details.
DeepSeek plans and pricing
| Plan | Price | What you get |
|---|---|---|
| Chat app | $0 | DeepThink reasoning and Search on web and mobile |
| API, deepseek-flash | $0.30 per 1M input tokens, $1.20 per 1M output (peak) | DeepSeek-V4.1-Flash-1017, 1M context, vision, half price off peak |
| API, deepseek-v4-pro | $1.32 per 1M input tokens, $3.96 per 1M output (peak) | DeepSeek-V4-Pro-0813, 1M context |
| API, Search and Fetch tools | $0.75 per 1,000 successful requests | Free through DeepSeek Harness, Claude Code or Codex until November 1, 2026 |
Frequently asked questions
Is DeepSeek free?
Yes. The DeepSeek chat app on the web, iOS and Android is free, with DeepThink reasoning, web search, file uploads and vision. There is no paid subscription. Only the developer API costs money.
How much does the DeepSeek API cost?
deepseek-flash costs $0.30 per 1M input tokens on a cache miss and $1.20 per 1M output tokens at peak, or $0.15 and $0.60 off peak. deepseek-v4-pro costs $1.32 and $3.96 at peak, or $0.66 and $1.98 off peak.
When is DeepSeek off peak?
Peak hours are 01:00 to 04:00 and 06:00 to 10:00 UTC, Monday through Friday, excluding Chinese public holidays. All other times, including weekends, are off peak at half price.
What is a cache hit on DeepSeek?
When the start of your prompt matches a recent prompt, DeepSeek bills those tokens at the cache hit rate: $0.006 per 1M at peak on deepseek-flash, about 50 times cheaper than a cache miss.
Does DeepSeek give free API credits?
DeepSeek's docs say billing draws from a granted balance first, then your topped up balance. The pricing page does not list a standard free credit amount.
How much do DeepSeek Search and Fetch cost?
$0.75 per 1,000 successful requests, free when used through DeepSeek Harness, Claude Code or Codex until November 1, 2026.
Is DeepSeek cheaper than OpenAI or Claude?
Yes, by a wide margin on per token rates. Anthropic lists Claude Sonnet 5.5 at $2 input and $10 output per 1M tokens; deepseek-flash at peak is $0.30 and $1.20.