Home / Bill repricer

Cloud Bill Repricer

Parse an AWS CUR-style CSV, GCP detailed-export JSON, or Azure cost-details CSV and see LLM token scenario costs alongside GPU-hour estimates. All parsing is local — no bill content is uploaded, stored, or written to browser storage.

Scenarios are not equivalent savings. Two source-qualified first-party target rate cards are shown below. All original catalog rows remain preserved and marked unqualified. Region, source modality, request context and workload equivalence hold all LLM scenarios. GPU-hour estimates for verified full-GPU AWS instance types are separate from whole-instance charges. No aggregate savings are shown. No currency conversion.

Choose CSV or JSON (max 5 MB)

Or paste CSV / JSON

Examples are invented test data, not real bills. Catalog snapshot: 26 Oct 2026; original price dates retained.

Source-qualified target rate cards

Checked 26 Oct 2026. Target-only rate cards — not proof of a comparable workload.

claude-sonnet-4-6 · Claude API (first-party)

$3/1M input · $15/1M output, USD

Base input/output token rates; cache operations separate. Global default routing, not an AWS region. See exact model context limits; request lengths must be verified.

5-minute cache write $3.75/M; 1-hour write $6/M; cache read $0.30/M. US-only inference is 1.1x all token categories. Partner-operated Bedrock and Google-managed prices are independent. Cache quantities and duration are not inferred.

Publisher pricing

Gemini 2.5 Flash · Gemini Developer API paid tier, not Vertex AI

$0.3/1M input · $2.5/1M output, USD

Text/image/video input; output includes thinking tokens. Region/data-residency equivalence not established. Publisher describes a 1M-token context window; source request lengths unverified.

Audio input is $1/M. Cache read text/image/video $0.03/M, audio $0.10/M, plus $1/M tokens/hour storage. Batch and Flex are $0.15/M input, $1.25/M output; Priority $0.54/M input, $4.50/M output. Those are separate tiers, not Standard. Tools and grounding costs excluded.

Publisher pricing

Nova 2.0 Lite · Amazon Bedrock (first-party AWS)

$0.33/1M input · $2.75/1M output, USD

Base input/output token rates; cache and other operations separate. us-east-1; region equivalence flagged for bill rows outside us-east-1. Source request lengths unverified; context pricing band not established.

Cache read is a separate standard usage type at $0.0825/M; cache write is free. Batch, flex, priority, and cross-region-global are separate SKUs with different rates and are not this standard-tier card. Custom-model SKUs carry different rates ($0.30/$2.50 per 1M) and are not covered here.

Publisher pricing

The original Google Gemini 2.5 Flash catalog row lists $0.15/M input, $0.60/M output and $0.0375/M cached input without verified tier or modality. It is retained as an unqualified catalog record, not used for a current Standard scenario.