Measured by us

Cost and latency — Claude Opus 5.5 on Direct API (global)

Results

Cost per task	0.037 USD lower is better
Time to first token (p50)	1.9 s lower is better
Output speed (p50)	58 tok/s higher is better

Run details

Model version: claude-opus-5-5
Endpoint: api.anthropic.com /v1/messages, streaming
Platform / region: Direct API / global
Dataset: latency-mix-100 (short / medium / long prompts)
Harness: GitHub @ a1b2c3d
Run cost: $3.70
Date: 23 Sept 2026

What was NOT tested

Fast mode
Priority tier
Regions outside the listed endpoint

Re-run it on your data

git clone https://github.com/mallimatla/uptodate.git && cd uptodate && git checkout a1b2c3d
npm install && npm run proof -- --suite cost-latency --model claude-opus-5-5 --platform direct --region global --dataset ./your-data