> ## Content Index
> Fetch the complete content index at: https://wire.fourthweb.ai/llms.txt
> Use this file to discover other available public pages before exploring further.

# DeepSeek Prices Web Search at $0.0007 Per Query While OpenAI Charges Per Token
- URL: https://wire.fourthweb.ai/deepseek-prices-web-search-at-0-0007-per-query-while-openai-charges-per-token/
- Published: 2026-10-18T11:31:51.000Z
- Updated: 2026-10-18T11:31:53.000Z
- Description: While OpenAI charges you per token to think, DeepSeek just priced web search at less than a tenth of a cent per query. DeepSeek launched a search API at $0.75 per 1,000 successful requests, letting AI agents pull live web data without per-token search overhead
- Author: Travis Wright
- Tags: AI Agent Economy, AI Agents, AI Infrastructure, OpenAI, China AI

**While** [**OpenAI**](https://wire.fourthweb.ai/tag/openai/) **charges you per token to think,** [**DeepSeek**](https://wire.fourthweb.ai/tag/china-ai/) **just priced web search at less than a tenth of a cent per query.**

### The Summary

- [DeepSeek launched a search API](https://api-docs.deepseek.com/guides/search/?ref=wire.fourthweb.ai) at $0.75 per 1,000 successful requests, letting [AI agents](https://wire.fourthweb.ai/tag/ai-agents/) pull live web data without per-token search overhead
- [DeepSeek v4.1 Flash 1017](https://openrouter.ai/deepseek/deepseek-v4.1-flash-1017?ref=wire.fourthweb.ai) now routes through OpenRouter, giving developers production access to China's fastest inference model
- The combo creates a new price floor for agentic web interaction: sub-penny search plus inference at a fraction of GPT-4 pricing

### The Signal

[DeepSeek's new search API](https://api-docs.deepseek.com/guides/search/?ref=wire.fourthweb.ai) charges $0.75 per 1,000 successful requests. Not per token. Not per completion. Per successful search query. That's 0.075 cents each time your agent needs to check the web. Compare that to embedding web search inside a model context window, where you're paying for tokens on both the query and the retrieved results. The pricing model matters because it changes what's economically viable to automate.

The timing lines up with [DeepSeek v4.1 Flash 1017's availability](https://openrouter.ai/deepseek/deepseek-v4.1-flash-1017?ref=wire.fourthweb.ai) through OpenRouter. The model itself has been generating serious Hacker News attention (427 points, 434 comments on [the Hugging Face announcement](https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017?ref=wire.fourthweb.ai)). Flash models are built for speed and cost, not reasoning depth. Pair cheap inference with cheap search and you've got infrastructure for agents that need to move fast and check facts constantly.

> "The combo creates a new price floor for agentic web interaction: sub-penny search plus inference at a fraction of GPT-4 pricing."

Here's what this unlocks practically:

- Customer service agents that verify account details, check order status, and pull support docs in real time without burning budget on bloated context windows
- Research assistants that can run dozens of parallel searches to triangulate information instead of relying on stale training data
- Monitoring bots that check competitor pricing, regulatory filings, or news mentions hundreds of times per day

The search API documentation shows DeepSeek is building for developers who want programmatic access, not just chat users who occasionally click a search button. That's the tell. They're pricing for volume because they expect agents, not humans, to be the primary customers. When search costs less than a tenth of a penny, you stop thinking about whether to search and start thinking about what patterns emerge when you search everything.

### The Implication

If you're building agents that need live data, DeepSeek just became your baseline comparison for search costs. The question isn't whether to use their API, it's whether anyone else can match this price without losing money. Watch for other providers to either match this pricing or explain why their search is worth 10x more.

For founders: cheap search infrastructure means your agent can be dumber and still useful. Instead of trying to cram everything into context or fine-tune for every edge case, you can build agents that know they don't know and look it up. That's a different product architecture than what most teams are building today.

### Sources

[Hacker News Best](https://api-docs.deepseek.com/guides/search/?ref=wire.fourthweb.ai)