GPT-6-Astra solves 49 of 50 problems on physics benchmark TheoPhysEval
What happened
On 15 October 2026, The Decoder reported that a new benchmark called TheoPhysEval had several commercial AI agents solve theoretical physics research problems without human hints. According to the report, GPT-6-Astra running in the OpenAI Codex environment solved 49 of 50 problems, a 98% accuracy rate, at an average cost of under $3 per problem and about $116 for the whole benchmark. The report also gives a figure of about $2.40 per problem. So far this is the only report on the progress, with no other sources or follow-ups.
Written by AI from the reports · updated 13 h ago
Timeline
Follow the reports to see the event from each side.
- The DecoderSelectedGPT-6-Astra solves 49 of 50 physics research problems at $2.40 apiece
On the TheoPhysEval benchmark, GPT-6-Astra solves 49 of 50 unpublished theoretical physics calculations as accurately as leading specialists. In some cases where the agent seemed to fail, the experts' own reference solutions turned out to contain errors. But the researchers say physics hasn't had a breakthrough like the ones in mathematics, because the agents lack "taste."
Heat of this event
Current heat 7·Peak in the comparable range 9（Oct 16 03:00）·Change over 24 hours in the comparable range –
The trend compares only the same accounts observed without a break, so its range can be narrower than the current heat count. Move the pointer or tap the chart to see each hour; with a keyboard, use the left and right arrow keys.