# ByteCosts AI Pricing Index

A machine- and human-readable index of source-derived AI model prices. Numbers come from a committed ByteCosts data snapshot; each row links to the provider source it was read from. Cite the source URL as the price origin.

_Last updated: 2026-07-19T21:42:52.054Z_

## Data coverage
- Providers indexed: 167
- Models tracked: 5695
- Models with pricing: 5296
- Source-derived rows in this index: 310
- Providers represented in this index: 38
- Evidence grades (source-derived rows): A = 92, A+ = 207, C = 10, D = 1

## Field definitions
| Field | Meaning |
| --- | --- |
| provider | Provider display name |
| model | Model name as published by the provider |
| input | Input price in the row's unit/currency (null if unknown) |
| output | Output price in the row's unit/currency (null if unknown) |
| unit | Pricing unit, e.g. USD_PER_1M_TOKENS |
| currency | ISO currency code, e.g. USD |
| context | Max context window in tokens (null if unknown) |
| sourceUrl | Provider page the rate was read from |
| lastCheckedAt | Snapshot regeneration timestamp (= generatedAt); when this index was last rebuilt from committed data, NOT a per-row re-verification |
| evidenceCheckedAt | ISO date this row's price was last verified against its provider source |
| confidence | Evidence level for the row (e.g. official) |

## Providers in this index (38)
Alibaba, Alibaba (China), Amazon Bedrock, Anthropic, Baseten, Cloudflare Workers AI, Cohere, Deep Infra, DeepSeek, Fireworks AI, Friendli, Google, Groq, Inference, Lilac, LucidQuery, MiniMax (minimax.io), MiniMax (minimaxi.com), NovitaAI, OpenAI, OpenCode Go, OpenCode Zen, OpenRouter, Perplexity, Perplexity Agent, routing.run, SiliconFlow, SiliconFlow (China), Together AI, v0, Vercel AI Gateway, Vertex, Vertex (Anthropic), Vivgrid, Vultr, xAI, Z.AI, Zhipu AI.

## Source and evidence policy
- Every row is sourced from a provider page (sourceUrl) and stamped with a lastCheckedAt date. Confidence reflects how directly the rate was evidenced.
- Rates are token-only and normalized to USD per 1M tokens unless the unit field states otherwise. They exclude volume discounts, commitments, and taxes.
- Data is gathered by local, manually-run scripts; the published site build is network-free and stores no provider API keys.

## Caveats
- `lastCheckedAt` is the snapshot regeneration time (equals the index `generatedAt`), not a per-row re-verification; use `evidenceCheckedAt` for the date a row's price was last verified against its source.
- Prices change frequently; always re-check the sourceUrl before relying on a number.
- Some rows have null input/output/context where the provider did not publish a comparable token rate.
- This index covers source-derived rows only; the full catalog (including aggregators and lab/list prices) lives on the site.

## Downloads
- JSON: https://bytecosts.com/ai-pricing-index.json (wrapped object: `{ generatedAt, source, count, rows }`)
- CSV: https://bytecosts.com/ai-pricing-index.csv (header row + one row per model)
- Full pricing index: https://bytecosts.com/pricing/