Amazon's Web Search Agent has secured the top position on the Search Arena leaderboard, surpassing OpenAI's GPT-5.6 Sol. This ranking is based on human votes rather than automated benchmarks, representing a significant shift in the competitive landscape for AI search agents.

The Search Arena platform, operated by Arena.ai, evaluates AI models by having users submit real-world queries and vote on the quality of the responses. Previously, in November 2026, OpenAI's GPT-5.6 Sol held the lead with a score of approximately 1255, while Amazon models were not among the top entries.

Amazon has been developing its search infrastructure for AI agents throughout 2026. The company launched Web Search in Amazon Bedrock AgentCore on June 16, 2026, and introduced the InvokeWebSearch API in November 2026 to allow developers direct access to search functions.

The search engine utilizes a combination of Amazon's web index and knowledge graph data to improve retrieval accuracy. This performance on the leaderboard serves as a marketing tool for Amazon, which charges $7 per 1,000 queries for its agent search services.

Search Arena was launched in early 2025, initially recording between 7,000 and 12,000 user votes. Early rankings were dominated by Google's Gemini-2.5-Pro-Grounding and Perplexity's Sonar models before OpenAI and Anthropic gained prominence in the field.