Not sure if this is reputable source but this lead...
# proj-agent
c
Not sure if this is reputable source but this leaderboard suggests Mistral Medium 3.5 does pretty well on SWE Bench Verified. llm-stats.com/benchmarks/swe-bench-verified relative to GLM-5, Qwen 3.7 Plus, etc. I notice it isn't on OpenHands Index and wondered about the backstory there...