MiniMax M3.1 Flash vs Opus 5.5 on 4 parallel agent jobs: 19:52 for ~$0.30 vs 10:57 for ~$4.52
Running the same four parallel agent jobs in native apps, M3.1 Flash took 19:52 at roughly $0.30 in estimated token cost versus 10:57 and at least $4.52 for Opus 5.5.
i’ve been testing the new minimax m3.1 flash against opus 5.5 high.. its big draw is coding + agent work per dollar.. so i gave both the same 4 jobs in their native apps. > 4 parallel agents each > both came back with bug fixes, an expense splitter, startup research + a cleaned csv > the whole run took 19:52 on minimax vs 10:57 on opus > estimated api token cost for everything.. ~$0.30 on minimax vs at least $4.52 on opus almost 2x the time.. roughly 1/15 the estimated api token cost. honestly.. for background agent work, i’d happily wait an extra 9 minutes for that cost difference... pretty impressed with how much it got done.. imagine leaving agents on your backlog overnight then wake up and review the work... at these prices.. minimax really cooked..
This post quotes another post: x.com/jumperz/status/2112124809148785117