OpenAI's newest model, GPT-5.6 Sol, outperformed Anthropic's Claude Opus in benchmark testing conducted during an early access period, according to a report from Crypto Briefing carried by Google News.
The report frames the result as a decisive one, saying GPT-5.6 Sol "crushes" the Claude Opus benchmark. The specifics of the testing — which benchmark was used, who ran it, and by what margin Sol came out ahead — are not detailed in the available source material.
Benchmarks are standardized tests that AI companies and independent evaluators use to compare how models perform on tasks such as reasoning, coding, and answering questions. Because the two companies are among the most closely watched developers of frontier AI systems, head-to-head comparisons draw heavy attention from developers and investors deciding which tools to build on.
It is worth noting that this account rests on a single early-access report, and benchmark claims made before a model's wider release often warrant independent confirmation.
Why it matters: The rivalry between OpenAI and Anthropic increasingly shapes which AI systems power the apps and services people use every day.