Claude Sonnet 5.5 Tops Coding Benchmarks but Burns 60% More Tokens Than Opus
The Batch (Andrew Ng) · rss · 2026-10-16
The Batch reviews Anthropic's new Claude Haiku 5.5 and Sonnet 5.5. Haiku 5.5 ($0.10/$0.01/$0.50 per M tokens) is the fastest model tested at 237 tok/s, ranking 13th on the Intelligence Index ahead of all fast models. Sonnet 5.5 ranks first on the Coding Agent Index (68) and second on Intelligence Index (56), but only at max reasoning — generating 193K output tokens per task, 60% more than Opus, with cost per task varying 18x across effort levels and TTFT averaging 430 seconds. Sonnet 5.5 is the first Sonnet with anti-distillation reasoning-extraction classifiers. Notably, it scores lower on FrontierCode at max (46.2%) than xhigh (52.1%) — test effort levels on your own tasks.
More from Models
- OpenAI's Decisions API hits GA: near real-time routing up to 10x faster than GPT-6 Luna — OpenAIDevs · 2026-10-23
- DeepLearning.AI weekly: GPT-6.1 hits 97.9% byte parity on COBOL migration — DeepLearningAI · 2026-10-23
- Marker's lite mode scores 75.1 vs 42.3 for Azure Read on OmniParseBench — VikParuchuri · 2026-10-23
- Datalab launches lite OCR at $0.50 per 1,000 pages, 20-30x cheaper than Azure, Google and AWS layout APIs — VikParuchuri · 2026-10-23
- LiveNerf's 29-day data finds no nerfing of Claude Opus 5.5 — TheOnlyVibemaster · 2026-10-23
- PHP Foundation reveals Anthropic's Claude cyber model scanned PHP, validated with OpenAI's Daybreak Red — saibharadwaj · 2026-10-23