Cheapest web search API for AI apps: DeepSeek, Parallel, Gemini
Web search for an AI app costs $0.75 per 1,000 calls with DeepSeek, $1 with Parallel and $10 with Claude or OpenAI. Gemini gives 5,000 free searches a month.
The main idea
If your AI app needs web search, the cost runs from $0.75 to $10 and more per 1,000 calls. DeepSeek launched a search API at $0.75 per 1,000 successful requests. Parallel Search has a fast mode at $1 per 1,000 requests on the turbo and fast tiers. Claude web search costs $10 per 1,000 searches, and OpenAI costs the same.
Google has a twist. Gemini gives you 5,000 free searches a month on the free tier. If you need about 5,000 searches a month, that is free, for something that would have cost about $50.
In the video I tried two paths. In the OpenRouter playground, I added DeepSeek V4 Flash and GPT OSS, asked for web search through Parallel on turbo mode, and the models got search added. Then I ran a Python script with Gemini grounding and asked who won the midterm election. It returned an answer based on actual search.
I did not give a final pick. I said DeepSeek is the cheapest I have seen, Parallel is nearly as cheap, and you should test that the search is accurate before you switch from the default OpenAI or Anthropic search.
Key points
- DeepSeek search API costs $0.75 per 1,000 successful requests.
- Parallel costs $1 per 1,000 requests on the turbo and fast tiers.
- Claude API web search and OpenAI web search each cost $10 per 1,000 calls.
- Gemini gives 5,000 free searches a month on the free tier.
- OpenRouter's playground can add Parallel search to models like DeepSeek V4 Flash and GPT OSS.
- Test that the cheaper search is accurate before you replace the default.
Chapters
- 0:35Price chart of search options
- 1:42Gemini free tier
- 2:01OpenRouter playground demo
- 3:33Gemini grounding from Python
- 5:13Claude and OpenAI search at $10 per 1,000
Save my free seat →