Opus 5.5 uses 15.6M tokens per task, output more than doubles to 333k
ArtificialAnlys · x · 2026-09-24
Artificial Analysis data: Opus 5.5 consumes 15.6M tokens per task, up 37% from Opus 5's 11.4M. Cached input rises from 10.9M to 14.6M, while output more than doubles from 137k to 333k tokens. This explains why per-task cost still rose 21% to $13.04 despite lower token prices.
Related event: Opus 5.5 burns 15.6M tokens per task, costs $13.04 despite price cut(3 posts)→
More from Models
- A private eval with a 0% completion rate for 3 years: no AI model can identify this flag — generativist · 2026-09-24
- Contrastive-LM org ships CLM-v0.1-8B model and Nemotron pretraining dataset on HF — _akhaliq · 2026-09-24
- Former OpenAI VP Brundage: ChatGPT web keeps resetting voice from Astra to Sol — Miles_Brundage · 2026-09-24
- Models now generate animated videos from scratch via HTML and Blender faster than predicted — xuanalogue · 2026-09-24
- Opus 5.5's safety classifier refuses to visualize another model's rollouts — eliebakouch · 2026-09-24
- Ex-OpenAI team launches Jev, a non-LLM model that outputs typed values with confidence scores, raises $40M — cephaloform · 2026-09-24