AsianFin — Aravind Srinivas, CEO of Perplexity, publicly praised the Qwen3-Coder model in a post on X, calling it “an amazing achievement” and declaring, “Open-source is winning big.”

His remarks follow a string of benchmark wins by Qwen3-Coder, which has outperformed Anthropic’s Claude 4 on key agent-focused evaluations such as SWE-Bench Multilingual, Aider-Polyglot, Spider2, and Mind2Web.

Qwen3-Coder is an open-source model developed in China and has quickly gained traction for both performance and affordability. Its API pricing is reportedly just one-third of Claude’s, making it an attractive option for AI developers and startups alike.

Given that Perplexity is one of Claude’s largest enterprise customers, Srinivas’s endorsement of Qwen3-Coder has sparked speculation across the AI industry. Analysts suggest his praise may signal the beginning of a broader shift from premium proprietary models to high-performing open alternatives, potentially setting off a new wave of “Qwen over Claude” adoption.