Skip to content

Latest commit

 

History

History

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 

README.md

DeepSWE harness comparison

Ten coding harnesses, one model, the same 113 tasks, one attempt each.

harness solved cost per task cost per solved issue mean time
senior-dev 62 of 113, 54.9% 22¢ 1x 54 min
mini-swe-agent 56, 49.6% 38¢ 1.9x 44 min
codex 51, 45.1% 37¢ 2.1x 46 min
pi 42, 37.2% 35¢ 2.4x 52 min
omp 31, 27.4% 50¢ 4.5x 49 min
opencode 30, 26.6% 50¢ 4.8x 48 min
kilo 30, 26.6% 48¢ 4.6x 54 min
claude-code 16, 14.2% 19¢ 3.4x 32 min
deepseek-harness 16, 14.2% 150¢ 26.6x 94 min
muse-code 3, 2.7% 12¢ 11.3x 16 min

senior-dev solved the most issues and paid the least for each one it solved: nearly 4x the issues claude-code solved, at about half the cost per solve of the next best harness.

Since then, on the same 113 tasks: 88 solved (77.9%, exact 95% CI 69.1% to 85.1%) with DeepSeek V4.1 Flash, and 78 (69.0%, 59.6% to 77.4%) with Kimi K3. Those runs are senior-dev alone, not a comparison.

Setup

Benchmark full DeepSWE set, 113 tasks, one seed per harness
Model deepseek/deepseek-v4-flash-0731 through OpenRouter
Verifiers official DeepSWE at 0b9fabb
Budget 3 h per task
Isolation four shards per harness, a dedicated OpenRouter key per harness
Ran nine harnesses on 2026-09-11; senior-dev on 2026-09-12

arms.csv has one row per harness: solved, reward rate, valid grades, invalid outcomes, mean F2P and P2P, OpenRouter spend, cost per task, mean agent seconds and model.

How the columns in the README are derived

  • solved: tasks the verifier passed, out of 113.
  • cost per task: billed OpenRouter spend divided by 113, so tasks without a verifier result stay in the denominator.
  • cost per solved issue: spend divided by tasks solved, shown as a multiple of senior-dev's (39.9¢).
  • mean time: mean agent wall time per task.

Limits

  • One seed per harness. senior-dev's 62 against mini-swe-agent's 56 is not a statistically resolved difference.
  • senior-dev departs from the sampling contract: the other nine sent temperature 1.0 and top-p 0.95, senior-dev sent neither, so provider defaults applied.
  • Five tasks produced no verifier outcome: codex 1, pi 1, omp 2, opencode 1. They count as unsolved.
  • omp, opencode, kilo, deepseek-harness, claude-code and muse-code did not record per-attempt cost; their spend is known at harness level only.
  • senior-dev ran under an earlier name for the binary; values are rewritten to senior-dev.