DeepSWE Measuring frontier coding agents(deepswe.datacurve.ai)
deepswe.datacurve.ai
DeepSWE Measuring frontier coding agents
https://deepswe.datacurve.ai/
1 comments
gpt-5.5xhigh leading benchmark, coincides with my recent experience. I've been opus 4.7 user but it burns tokens so quickly, so gave gpt-5.5xhigh (via codex) a try, quality was similar (if not better), and tokens lasted a lot longer.