Engadget Analysis Reveals Diverging Paths for Anthropic’s Claude and OpenAI’s ChatGPT

A comparative analysis published by Engadget has highlighted significant distinctions between Anthropic’s Claude and OpenAI’s ChatGPT, focusing on accuracy, hallucination rates, and user demographics. The report utilises data from the AA-Omniscience benchmark suite to evaluate the platforms, revealing that while flagship models show marginal differences in accuracy, mid-tier models and error rates tell a more complex story. Claude’s flagship Fable 5 model achieved a 61 percent accuracy score compared to ChatGPT’s GPT 5.6 Sol at 59 percent, though the report notes this difference is rarely perceptible in everyday usage.

The disparity becomes more pronounced when examining mid-tier models and hallucination metrics. ChatGPT’s 5.6 Terra model outperformed Claude’s Sonnet 5 in accuracy, scoring 46 percent against 38 percent. However, Claude demonstrated a substantial advantage in reliability, recording a hallucination rate of 37 percent for its mid-tier model compared to ChatGPT’s 85 percent. For flagship models, Claude’s Fable 5 scored 55 percent on the hallucination benchmark, significantly lower than ChatGPT’s 5.6 Sol, which scored 89 percent. The analysis also noted that OpenAI’s latest models have seen an increase in hallucinations compared to previous iterations, with GPT-4o recording a rate of 38 percent.

Usage patterns indicate a sharp divergence in how the tools are deployed. Data from the Anthropic Economic Index report from March 2026 shows that 45 percent of Claude conversations were work-related, with 42 percent used for personal purposes. In contrast, an OpenAI report indicated that 70 percent of ChatGPT usage is non-work-related. This professional inclination is supported by features such as Claude Cowork and the ability to create instruction bundles called skills, which can be invoked across chat and coding environments. ChatGPT’s equivalent features, such as Sites, are more restricted, targeting internal business use within the Codex environment and lacking the ability to pull live data via Model Context Protocol connectors.

Commercial strategies and feature implementations further separate the platforms. ChatGPT has introduced advertisements to its free and Go tiers, a move described by Engadget as a potential precursor to a degraded user experience given OpenAI’s financial pressures. Claude maintains an ad-free free tier but offers broader access to mid-tier models. Both providers offer similar paid structures, with monthly plans at $20, $100, and $200, though ChatGPT uniquely offers an $8/month tier that includes ads. Feature-wise, ChatGPT retains an edge in voice interaction naturalism and photorealistic image generation, whereas Claude focuses on rendering code snippets, HTML websites, and interactive diagrams.

User migration trends have been influenced by recent geopolitical developments. In February 2026, Anthropic refused to supply the US Department of War with technology for mass domestic surveillance and autonomous weapons, leading to its designation as a supply chain risk. OpenAI subsequently signed a similar deal with the US government. This ethical stance by Anthropic prompted a surge in downloads for Claude as users migrated from ChatGPT, perceiving Anthropic as the more ethically aligned provider. The analysis concludes that while both platforms continue to evolve, their differing approaches to safety, monetisation, and utility cater to distinct user bases.