DeepSeek API pricing (Nov 2026)
deepseek-flash peak prices are $0.30 cache-miss input, $0.006 cache-hit input, and $1.20 output per million tokens; deepseek-v4-pro peak prices are $1.32 cache-miss input, $0.044 cache-hit input, and $3.96 output per million tokens (off-peak is half for both). The full table is on the official pricing page. See the off-peak hours guide. For cross-vendor list prices see the cost-effective LLM API comparison, and for benchmarks, speed and usage against Claude, Gemini, GPT and Kimi see DeepSeek vs frontier models. See also the model name migration guide, Search & Fetch guide, and API error codes.
API list prices & model specs (official)
Peak and off-peak USD per 1M tokens. Off-peak is half of peak; peak hours are listed on the pricing page.
Is the DeepSeek API free?
No. DeepSeek bills model usage per token at the published list prices, and Search/Fetch per successful request. The Models & Pricing page says fees are deducted from your topped-up balance or granted balance (granted balance is used first when both are available). This directory does not publish grant amounts or signup bonuses — check the DeepSeek Platform for your account. When balance runs out, the API returns HTTP 402 (Insufficient Balance) per the error codes page.
Price changes (sourced)
Only dated changes from the official API change log entries mirrored on /news. Current list prices are always on the live pricing page.
| Date | What changed | Source |
|---|---|---|
| 2026-10-17 | Search and Fetch APIs opened to all accounts; $0.75 per 1,000 successful requests each. News | API change log |
| 2026-09-10 | DeepSeek-V4.1-Flash release; API prices were reduced (V4 Pro billing unchanged until further notice). News | API change log |
| 2024-08-02 | Disk context caching launched; cache-hit input pricing reduced (see current Models & Pricing table). News | API change log |
Search & Fetch API
DeepSeek opened Search and Fetch to all API accounts on 17 October 2026 (per the API change log). Billing is per successful request, not by tokens or by how many results Search returns.
Search
POST /v1/tools/search
$0.75 per 1,000 successful requests
Fetch
POST /v1/tools/fetch
$0.75 per 1,000 successful requests
| API | Endpoint | Price (per 1,000 successful requests) |
|---|---|---|
| Search | POST /v1/tools/search | $0.75 USD |
| Fetch | POST /v1/tools/fetch | $0.75 USD |
Apps, API Integration & Economics
Access official portals, OpenAI SDK compatibility configurations with dynamic reasoning effort parameters, sourced model identifiers, and interactive cost comparison models.
Official Applications & Developer Portals
DeepSeek Chat (Web & Mobile)
Official interactive web and mobile client featuring free access to V4.1-Flash, DeepSeek-R1, and Janus-Pro multimodal image tools.
DeepSeek Developer Platform
Provision API keys, inspect token billing, track cache hit ratios, and test in the interactive API playground.
DeepSeek Open Source Hub
Official GitHub organization hosting model weights, FlashMLA, DeepEP, DeepGEMM, DualPipe, 3FS, and Engram memory repositories.
Base URL and headers
DeepSeek offers an OpenAI-compatible API gateway. Point your SDK to https://api.deepseek.com and pass your bearer token to immediately run V4.1-Flash with dynamic reasoning effort.
import OpenAI from 'openai';
// Initialize OpenAI client pointing to DeepSeek API
const openai = new OpenAI({
baseURL: 'https://api.deepseek.com',
apiKey: process.env.DEEPSEEK_API_KEY || '<YOUR_KEY>',
});
async function main() {
const stream = await openai.chat.completions.create({
model: 'deepseek-flash',
messages: [
{ role: 'system', content: 'You are an expert systems engineer.' },
{ role: 'user', content: 'Design a lock-free queue with zero-copy buffer in C++20.' },
],
stream: true,
reasoning_effort: 'high',
extra_body: {
thinking: { type: 'enabled' },
},
});
for await (const chunk of stream) {
const delta = chunk.choices[0]?.delta as any;
// Inspect dynamic reasoning thoughts if available
if (delta?.reasoning_content) {
process.stdout.write(delta.reasoning_content);
} else if (delta?.content) {
process.stdout.write(delta.content);
}
}
}
main();Interactive Pricing Calculator & Cache-Hit Economics
| Provider & Model | Cache Miss / 1M | Cache Hit / 1M | Output / 1M | Est. Monthly Cost | Source |
|---|---|---|---|---|---|
DeepSeek-V4.1-Flash (peak)DeepSeek | $0.30 | $0.006 | $1.20 | $14.51 USD | API pricing |
DeepSeek-V4-Pro-0813 (peak)DeepSeek | $1.32 | $0.044 | $3.96 | $51.37 USD | API pricing |
Model Identifiers for Production API Calls
deepseek-flashCurrent flash API model. Legacy names deepseek-v4-flash and deepseek-v4-flash-vision-exp are still accepted and are served by V4.1-Flash at the flash price. 1M context, vision supported.
deepseek-v4-proAPI checkpoint DeepSeek-V4-Pro-0813. 1M context, max output 384K, vision not supported. Weights card: 1.6T total, 49B activated.
HTTP REST API Endpoints
OpenAI-compatible chat completions. Current model names on the pricing page are deepseek-flash and deepseek-v4-pro. Thinking mode uses thinking.type plus reasoning_effort (low, high, or max). The old extra_body.effort field is not the documented control.
Lists active foundation models, reasoning checkpoints, and multimodal endpoints with context window permissions.
Returns real-time account balances in USD and CNY, cached prefix savings metrics, and monthly grant status.