Sonnet 5.5 is a token-hungry monster: 193k output tokens, 2x GPT-6's cost per task
haider1 · x · 2026-09-29
Cost analysis of Claude Sonnet 5.5 on the AA Intelligence index shows it's a token-hungry monster: Sonnet 5.5 max outputs 193k tokens per task vs Opus 5.5 max's 119k, Sonnet 5 xhigh's 65k, GPT-6 sol max's 31k, and GPT-6 astra max's 27k — 62% more than Opus 5.5 and over 6x GPT-6 sol. On cost, Sonnet 5.5 max runs $7.60/task vs GPT-6 astra max's $3.26 and Sonnet 5 xhigh's $2.87 — more than 2x Astra's per-task cost.
Related event: AA Benchmarks: Sonnet 5.5 Nears Opus 5.5 but Sets Output Token Record(9 posts)→
More from Models
- Altman teases 'We have found a new thing' ahead of DevDay — signulll · 2026-09-29
- Tip: don't use Sonnet 5.5 at max — it's dumber and far more expensive; high hits the best cost per task — airesearch12 · 2026-09-29
- Vibe Check: Sonnet 5.5 Is a More Capable Partner Than Sonnet 5, If You Keep a Hand on the Wheel — kieranklaassen · 2026-09-29
- Third-party benchmark shows Claude Sonnet 5.5 crushing GPT-6 sol, nearly matching GPT-6 Astra — rickasaurus · 2026-09-29
- Qwen3.8 Flash hits 74 tok/s single-stream, 212 tok/s aggregate on one DGX Spark — open vLLM recipe — DimeRhyme · 2026-09-29
- Anthropic says Sonnet 5.5 at low/medium effort beats Sonnet 5's best scores at ~1/10 the cost — airesearch12 · 2026-09-29