Coding agent index shows max effort costs 6x for fewer points; Gemini 4 Neon wins on value

Arindam_1729 · x · 2026-11-20

Citing Artificial Analysis' Coding Agent Index, Arindam argues effort is a budget dial, not a quality dial: Claude Fable 5.2 in Claude Code scores 58 at $7.51/task on high and 58 at $15.47 on xhigh, but only 56 at $44.22 and 2.2 hours/task on max — 6x the cost for 2 fewer points. Meanwhile Gemini 4 Neon in Antigravity CLI scores 59 at $6.86/task, beating Claude Opus 5.5 max ($12.61). His advice: default to max isn't safe; pick the lowest effort that clears your tasks and spend the savings on verification. Caveat: these are pay-per-token API costs; subscription economics differ, though the 2-hour runtimes remain.

Original post →

More from coding & agent

coding & agent channel →