Hi all! Quick question: the OpenHands Index gives one harness (OpenHands) x many models with per-instance results. Are there known resources doing the same for other harnesses — per-instance results or trajectories across different models under a fixed harness (on SWE-bench Verified/Multimodal, SWT-Bench, GAIA)?

Hey @elegant-sandwich-94187 I suppose yes. Are you interested in a particular harness?

Hey @elegant-sandwich-94187 did you try clicking on the "alternative agents" tab?

Hey @modern-fireman-99755 and @limited-student-10747 Yeah, I checked out the Alternative Agents tab! I was wondering if there are more comprehensive cross-model sweeps available for other harnesses (like Aider, Terminus-2, etc.). I'm hoping to find a wider matrix or raw per-instance trajectories and associated costs across different (model, harness) pairs. I am guessing that Codex or Claude-CLI only work with their own respective models.

Hey @elegant-sandwich-94187, we don’t have that at the moment. The main barrier is mostly just the cost and our time.