HLE-Diamond: Claude Opus 5.5 gains 26 points at max effort — at 35x the per-task cost
ArtificialAnlys · x · 2026-10-30
Artificial Analysis published full HLE-Diamond benchmark results, including scores, token usage and cost per task.
Key findings on effort settings:
- Claude Opus 5.5: scores jump from 42% (low effort) to 68% (max), a 26-point gain, but at roughly 35x the cost per task
- GPT-6 Astra: 45% → 64% (+19 points) at about 11x the cost
The data shows higher effort settings substantially boost HLE-Diamond scores, but with steeply rising marginal costs that vary widely between models. Full score, token and cost breakdowns are available on their site.
More from Models
- Heavy user calls OpenAI's 6.1-terra the laziest model they've ever used — lxfater · 2026-10-30
- Rumors swirl: OpenAI quota cut next week as Dots' first-month free allowance promo ends — lxfater · 2026-10-30
- Grok Bot on Windows plagued by flashing messages, bot spam and voice issues — IndraVahan · 2026-10-30
- Altman: OpenAI Almost Ditched Token Pricing for an AWS-Style Bill, Users Refused — menhguin · 2026-10-30
- User math: $1 of Son costs 7x Opus on $20 plan, but tasks only 1.3x — alvelda · 2026-10-30
- Claude gets the right answer but can't prove time complexity — geoffreyirving · 2026-10-30