Hacker News
new
top
best
ask
show
job
DeepSWE Benchmark Results for GPT 5.6
(
deepswe.datacurve.ai
)
2 points
by
theanonymousone
8 hours ago
1 comment
tomerlir
8 hours ago
borderline better
theanonymousone
4 hours ago
I wouldn't call it borderline. If we take the chart seriously (if), it means for every budget, the best performance is achieved by a 5.6 variant.