Chinese Tech Firm Launches In-House AI Search Engine at Fraction of Western Competitor Costs

2026-10-19 companies

Hangzhou, Sunday, 18 October 2026.
DeepSeek has introduced its upgraded AI model and proprietary web search API at $0.75 per 1,000 calls—undercutting rivals like OpenAI and Claude by over 90%.

Benchmark Improvements and Model Efficiency

The newly released DeepSeek-V4.1-Flash-1017 model demonstrates measurable gains in performance metrics compared to its predecessor. According to benchmark comparisons shared with the announcement, the updated model scored 77.0 on the DeepSWE v1.1 benchmark, an increase from 74.2 for the previous version [1][5]. This represents a performance improvement of 3.774 percent on this specific metric [1][5]. Additionally, the model achieved a score of 60.5 on Automation-Bench V1.0.6, compared with 54.8 previously, indicating enhanced capabilities in automated task execution [5]. The company describes this update as a “Pareto improvement,” signifying gains in capabilities and efficiency without a corresponding increase in expense [5]. The model is optimized for tools such as DeepSeek Harness, Tencent CodeBuddy, and WorkBuddy, utilizing reinforcement learning with self-feedback to improve end-to-end project development [5].

API Availability and Pricing Structure

Concurrent with the model update, DeepSeek opened its Search and Fetch APIs to all users on Saturday, 17 October 2026 [6]. The services are powered by the company’s in-house search engine, transitioning away from third-party providers used throughout 2025 [6]. Each Search or Fetch API endpoint costs $0.75 per 1,000 calls, a rate confirmed across multiple documentation sources [5][7]. The Search API returns a ranked list of structured results, while the Fetch API retrieves web page content formatted as Markdown [2][3]. Technical documentation specifies that the Search query must contain 1 to 128 Unicode characters, and the Fetch URL must be at most 8192 characters [2][3]. Users are limited to 5 queries per second per account under the current rate limits [6]. Notably, Search and Fetch API calls via DeepSeek Harness, Claude Code, and Codex are free until 00:00 Beijing time on 1 November 2026 [6].

Market Position and Infrastructure Scale

The pricing strategy positions DeepSeek significantly below Western competitors in the search API market. Comparative market data indicates competitors such as OpenAI and Claude charge approximately $10 per 1,000 calls, while Perplexity Fast charges $1 and Brave charges $5 [4]. At the DeepSeek rate of $0.75, the cost is 92.5 percent lower than the $10 price point observed for major rivals [4]. Behind the service, DeepSeek’s infrastructure includes a Fire-Flyer cluster with 180 storage nodes and over 20 PiB capacity [6]. The search engine currently indexes hundreds of billions of web pages and processes hundreds of millions of search requests daily [6]. This expansion intensifies price and performance competition for Western tech giants and presents macroeconomic implications for enterprise software spending [1]. Investors and users are advised to monitor DeepSeek changelog entries and pricing documentation for changes against the 18 October 2026 baseline [4].

Sources

Artificial Intelligence DeepSeek