GPT-6-Astra solves 49/50 problems on physics benchmark TheoPhysEval
Understand the story
On October 15, 2026, The Decoder reported that researchers launched a new theoretical physics benchmark called TheoPhysEval, designed to test routine research-level problems. On the benchmark's 50 questions, OpenAI's model GPT-6-Astra (running in the Codex environment) solved 49 of them without any prompting, i.e., 49 out of 50. The report also noted that the cost of solving a single problem is about $2.40. Public reporting so far includes only this result, with no further details about the benchmark, other models' performance, or a response from OpenAI.
AI overview of reporting · Updated 3 days ago
Reporting timeline
Explore different aspects of the story through its reporting.
- The DecoderGPT-6-Astra solves 49 of 50 physics research problems at $2.40 apiece
Researchers introduced TheoPhysEval, a benchmark of routine theoretical physics research problems, and OpenAI's GPT-6-Astra (running in the Codex environment) solved 49 of 50 problems without human hints.
Story heat trend
Current heat3·Comparable peak10（10/16 00:00）·Comparable change over 24 hours -51%
The trend compares the same continuously observed sources, which may cover fewer sources than the current heat score. Hover or select the chart to inspect each hour; use the left and right arrow keys to navigate.