Claude Sonnet 4.5 Surpasses OpenAI's GPT-5 Codex in Coding Performance and Latency, Developer Claims

A recent social media post by developer Kyle Corbitt has ignited discussion within the AI community, asserting that Anthropic's newly released Claude Sonnet 4.5 consistently delivers superior results and lower latency for coding tasks compared to OpenAI's GPT-5 Codex. Corbitt stated in his tweet, "> I'm getting consistently better results with Claude Code (sonnet 4.5) than Codex (gpt-5-codex). The lower latency also significantly helps the user experience." This claim highlights a competitive edge for Anthropic in the rapidly evolving landscape of AI-powered code generation.

Anthropic officially launched Claude Sonnet 4.5 on September 29, 2025, positioning it as a frontier model specifically designed for agentic work, coding, and complex computer operation tasks. The model boasts state-of-the-art performance, achieving a 77.2% accuracy on the SWE-bench Verified coding benchmark and 61.4% on OSWorld, an evaluation for computer tasks. Key features include checkpoints within Claude Code, an Agent SDK for building autonomous agents, and enhanced context management capabilities.

While OpenAI announced its GPT-5 model in August, details surrounding a specific "GPT-5 Codex" remain less clear, with some reports indicating it as a high-performance coding variant. According to a market analysis, Claude Sonnet 4.5's performance on SWE-bench Verified is "slightly ahead of GPT-5 Codex at 74.5%," corroborating Corbitt's observation. However, the same analysis notes that the pricing for GPT-5 and GPT-5 Codex is reportedly $1.25 per million input tokens and $10 per million output tokens, making it more cost-effective than Sonnet 4.5's $3/$15 per million token rate.

The emphasis on lower latency by Corbitt underscores a critical factor for developers, as faster response times can significantly improve user experience and workflow efficiency in real-time coding environments. Claude Sonnet 4.5 has been integrated into platforms like Amazon Bedrock and GitHub Copilot, making its advanced coding capabilities widely accessible to developers. Anthropic claims Sonnet 4.5 is "the most aligned frontier model" they have released, showing substantial improvements in various aspects compared to previous Claude models.