Claude Sonnet 5.5 Efforts Span 18x Price Range: $0.41 to $7.60 Per Task on AA Index
ArtificialAnlys · x · 2026-09-29
Artificial Analysis published a comparison page for Claude Sonnet 5.5's five reasoning-effort variants:
- Intelligence Index: max effort scores 56, with xhigh at 52, high 47, medium 41, low 36.
- Cost per task: low effort is cheapest at $0.41, max reaches $7.60 — up to an 18x spread across efforts.
- Specs: text+image input, 1M context window (1500 A4 pages); speed metrics pending.
The data quantifies how max effort buys performance with massive reasoning volume: 18x the cost for roughly a 20-point intelligence gain.
More from Models
- OpenAI Teases Major Announcement With Cryptic 'Get Ready' Post — OpenAI · 2026-09-29
- Claude Sonnet 5.5 lands in Rork, generating output over 30% faster than Sonnet 5 — rudrank · 2026-09-29
- AI models' viral demo appeal is underpriced, argues founder who picked Opus over better-per-dollar rival — gabriel1 · 2026-09-29
- Open-source 149M LateOn decision model ditches classifier heads for late interaction — antoine_chaffin · 2026-09-29
- Opus 5.5 Is Deliberately Squeezing Budget Users, VC Argues — 'There Is No Single Model' — StewartalsopIII · 2026-09-29
- Tesla Roadster event pushed to Oct 15 over weather, with OpenAI set to announce tomorrow — Scobleizer · 2026-09-29