How the number is calculated

Formulas

api_month = queries_per_day × days × (input_tokens/1e6 × input_price
                                    + output_tokens/1e6 × output_price)
hosting_month = vps_month + other_month
total_month = hosting_month + api_month + hardware_depreciation
per_query = total_month / (queries_per_day × days)
break_even_queries = ceil(20 / per_query)

In local mode, API cost is set to 0 and hardware cost is spread over the depreciation months. If you leave hardware at 0, only hosting is counted.

Defaults

• VPS: Hetzner CAX11 reference $4.13/month; Railway reference ~$10/month. Both are references, not quotes.
• Model: Claude Sonnet 5.5 reference $2.00 per 1M input tokens, $10.00 per 1M output tokens.
• Usage: 100 queries/day, 2,000 input tokens and 500 output tokens per query, 30 days/month.

Sources and check date

• Vane official GitHub README: install, Docker, SearXNG, model configuration.
• Railway Vane template: single container, port 3000, ~$10/month reference.
• ossreplace Vane cost page: Hetzner CAX11 from about $4.13/month (third-party reference).
• Hetzner Cloud pricing: the official checkout is the source of truth.
• Hostinger Vane page: hosting-plan reference only; not a default preset (checked 2026-10-21).
• Claude Sonnet 5.5 reference: a model marketplace listing.
• Anthropic official pricing: the model vendor’s own price page; defaults are checked against it.

All prices checked 2026-10-21. Prices, taxes, exchange rates, and plans change; these numbers are for decisions, not invoices.

What it cannot tell you

• Taxes, electricity, bandwidth, backups, monitoring, or your maintenance time.
• Model quality, latency, context limits, and rate limits.
• Real CPU/disk performance across VPS providers; CAX11 is ARM, so check image compatibility before buying.