Measured by us

Cost and latency — Claude Opus 5.5 on AWS Bedrock (ap-south-1)

Results

Cost per task	0.039 USD lower is better
Time to first token (p50)	1.4 s lower is better
Output speed (p50)	52 tok/s higher is better

Run details

Model version: claude-opus-5-5
Endpoint: Bedrock Mantle Messages API (preview), streaming
Platform / region: AWS Bedrock / ap-south-1
Dataset: latency-mix-100 (short / medium / long prompts)
Harness: GitHub @ a1b2c3d
Run cost: $3.90
Date: 23 Sept 2026

What was NOT tested

Batch inference (not available in region)
Quota limits above preview defaults

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 bedrock --region ap-south-1 --dataset ./your-data