Got some feedback from the Evoclaw team that compa...
# proj-agentc
chilly-france-44439
05/07/2026, 9:02 PMGot some feedback from the Evoclaw team that compares OpenHands w/ claude code using kimi-k2.6 and GLM-5; and it seems we had some agent performance issues 👀
https://github.com/OpenHands/software-agent-sdk/issues/3112
chilly-france-44439
05/07/2026, 9:03 PMcc @limited-student-10747 @strong-ice-84145 @modern-fireman-99755
l
limited-student-10747
05/07/2026, 9:12 PMah, interesting, we can try to repro with our benchmarks at some point
m
modern-fireman-99755
05/07/2026, 9:57 PMAre the traces avaible?
modern-fireman-99755
05/07/2026, 10:02 PMAh ok I read the issue now :)
modern-fireman-99755
05/07/2026, 10:10 PMI will give a look inside their logs and/or traces
Good excuse to get my hands dirty with EvoClawBench ;)
🙌 1
c
chilly-france-44439
05/08/2026, 4:38 AM😄 i also set up my own reproduction stack, it might have to do with the weird way they are using the SDK though
https://github.com/EvoClaw-Bench/EvoClaw/blob/main/harness/e2e/agents/openhands.py
chilly-france-44439
05/08/2026, 4:38 AMi was poking around and wonder if we can add it to our benchmarks
https://github.com/OpenHands/benchmarks/pull/705
chilly-france-44439
05/08/2026, 4:38 AMmaybe eventually in the next verision of OH Index
2 Views