Musk: Grok Code Fast 1 surpasses Claude Sonnet, tops OpenRouter leaderboard.

Musk: Grok Code Fast 1 surpasses Claude Sonnet, tops OpenRouter leaderboard.

Elon Musk’s xAI Holdings Corp. has made significant strides in the AI coding market with the release of Grok Code Fast 1. According to the latest benchmarks, Grok Code Fast 1 has surpassed Claude Sonnet, a model developed by Anthropic, and now tops the OpenRouter leaderboard [1]. This achievement underscores xAI’s commitment to delivering high-performance, cost-effective AI solutions for developers.

Grok Code Fast 1 is designed to be a speedy and economical coding agent, addressing the common frustration of sluggish performance in agentic coding workflows. The model was engineered from the ground up with a new architecture, focusing on speed and efficiency. It supports multiple programming languages including TypeScript, Python, Java, Rust, C++, and Go. According to internal benchmarks, Grok Code Fast 1 can execute at approximately 160 tokens per second, outperforming competitors like OpenAI’s GPT-5, which averages around 50.1 tokens per second [1].

The model's responsiveness is further enhanced by prompt caching optimizations, achieving hit rates above 90% when used with launch partners. This focus on performance is paired with an aggressive economic strategy, with pricing at $0.20 per million input tokens, $1.50 for output, and $0.02 for cached inputs, significantly undercutting many rivals [1].

In addition to its speed and cost-effectiveness, Grok Code Fast 1 has demonstrated strong performance in real-world coding tasks. It achieved a 70.8% score on the full subset of SWE-Bench-Verified, a human-validated evaluation of AI models’ ability to solve real-world software engineering problems [1]. This score places it firmly in the top tier of a crowded field, where the title of “best coding model” changes hands frequently.

xAI’s strategic move to release Grok Code Fast 1 for free for a limited time through various AI-enabled coding platforms, including GitHub Copilot and Cursor, is expected to drive adoption and further validate its approach. The company’s plans to deliver consistent updates and improvements on a daily basis also indicate a commitment to maintaining its competitive edge in the rapidly evolving AI coding market.

References:
[1] https://siliconangle.com/2025/08/29/elon-musks-xai-releases-agentic-coding-model-aimed-high-speed-affordability/