2 Comments
User's avatar
Casey Spaulding's avatar

Not sure what those benchmarks are testing but GPT 5 is far behind Sonnet even! In coding at least.

Rich Holmes's avatar

Thanks for sharing. These are from the SWE-Bench verified test, but you're right: early anecdotal experiences from developers since GPT-5 was launched show that it still doesn't seem to beat Anthropic in coding despite this. 🤷‍♂️