how the numbers get here
Methodology
Primary sources only
Every rate on this site is read from the provider's own published price list — the pages linked from each provider profile. We never source prices from other comparison sites, forums, or screenshots. Each model record carries the URL it was read from.
What we record
- Standard pay-as-you-go rates in USD per 1M tokens: input, cached input where offered, and output.
- Monthly subscription prices on monthly billing, with the effective annual rate alongside where the vendor publishes one.
- Context windows and output limits where officially documented.
What we deliberately leave out
- Batch, off-peak, and enterprise-negotiated discounts — they vary by contract and would make rows incomparable.
- Promotional prices are recorded at their published value; when a provider publishes a scheduled increase we note it in the changelog.
- Numbers we cannot confirm on an official page. A missing cell means “not published”, never “we guessed”.
Tier labels are editorial
“Frontier / Mid / Budget” is our judgment based on price and the provider's own positioning, not an official designation. Use it to scan, not to decide.
Verification cadence
Rates were last verified on 2026-08-31. Providers reprice without notice, so always confirm on the official page before committing a budget. Every change we record lands in the changelog.
Independence
No provider pays for placement, ordering, or color on this site. Sorting is by price; the heat map is a pure function of the numbers.
Providers currently tracked
- OpenAI — official price list
- Anthropic — official price list
- Google — official price list
- DeepSeek — official price list
- xAI — official price list
- Mistral — official price list
- Moonshot AI — official price list
- Alibaba Qwen — official price list
- Z.ai — official price list
- MiniMax — official price list
- Cohere — official price list
- Perplexity — official price list
- Groq — official price list
- Together AI — official price list