Tokenomics

Real AI cost. Not vendor list price.

Cloud pricing live-scraped from each provider weekly. Self-hosted £/1M-tokens computed from real-world tokens-per-second on real GPUs — anonymously aggregated from production agents, not vendor specs. Capability, ESG, and ROI alongside.

Choose use case

Support tickets

Summarise and classify support tickets

1000 tickets per month default

CRM queries

Natural language queries over CRM pipeline

500 queries per month default

Document summarisation

Summarise documents (contracts, reports)

200 documents per month default

AI Chat (conversational)

General conversational AI: drafting, research, Q&A, brainstorming

2000 messages per month default

Learn more

Why these numbers?

Pick a use case above and we'll show how the volume default was chosen.

Refine your scenario

Volume (per month)

Number of requests per month. Adjust if you know your usage.

Advanced: adjust token assumptions

Input tokens per request

Tokens of context sent to the model. ~4 characters per English token, so a 500-word prompt is ~670 tokens.

Output tokens per response

Tokens the model generates in response. Adjust if you know your usage.

Self-hosted assumptions (amortisation, electricity, GPU utilisation)

Hardware amortisation

1 yr 3 yr 5 yr

How long you spread the GPU's purchase price across.

Electricity rate (p/kWh)

UK domestic standard: 26.35 p/kWh.

GPU utilisation (%)

Share of the day the GPU is actively serving inference. 100% is unrealistic; 70% reflects a well-loaded production deployment.