Price per 1M logo

Every model we track, one page each

The dataset behind this site contains 1,657 endpoints from 32 providers, but many of those rows are the same model resold by a different platform. Rather than publish a near-identical page for every listing, listings are grouped by model name: each model gets one page that shows what every platform charges for it, which is a comparison no single provider's pricing page can give you.

This page lists the 45 canonical model pages in the curated index, grouped by provider. The 897 remaining model groups are reachable from the pages below but are not offered to search engines — the reason is explained on the About page. Data refreshed 2026-10-28 12:57 UTC.

Canonical model pages
45
one per tracked model
Priced chat models
1,557
both input and output published
Providers
32
own APIs and hosting platforms
Platform listings behind them
1,657
grouped onto the pages below

Which model should you open first?

Cheapest chat APIs
Ranked by blended cost for a typical input/output mix.
Cheapest for a support bot
Weighted for high-volume, cache-heavy short answers.
How to budget for one
Turn a per-1M-token rate into a monthly number.

OpenAI

17 models · provider detail

OpenAI bills its own API per million tokens, with a separate, lower rate for cached input and tiered limits that rise with account spend.

Alibaba DashScope

5 models · provider detail

Alibaba prices Qwen models per million tokens in the DashScope API, with cached-input discounts and regional billing.

DeepInfra

4 models · provider detail

DeepInfra hosts open-weight models per million tokens, with cache-read discounts on some endpoints and no minimum spend.

DeepSeek

4 models · provider detail

DeepSeek prices per million tokens with an off-peak discount window, so the same model can cost materially less at different hours of the day.

Google Gemini

3 models · provider detail

Google prices Gemini per million tokens in its own API, with a cheaper cached-input rate and free-tier quotas that sit outside the paid rates listed here.

Moonshot AI

3 models · provider detail

Moonshot AI publishes per-million-token rates for Kimi models, with long-context requests billed at a higher tier.

Perplexity

3 models · provider detail

Perplexity charges per request or per token depending on the endpoint, and its Sonar models are billed through the same account as the consumer product.

Anthropic

2 models · provider detail

Anthropic prices per million tokens directly, with prompt-cache reads and writes billed separately from fresh input, and long-context requests billed at premium tiers.

Hyperbolic

2 models · provider detail

Hyperbolic prices per million tokens for open-weight models on its decentralised GPU pool.

MiniMax

1 model · provider detail

MiniMax bills per million tokens through its own API, with separate rates for its chat and audio endpoints.

Nscale

1 model · provider detail

Nscale bills per million tokens for open-weight models served from European data centres.

Why the index is curated rather than exhaustive

A price index is only useful if two rows are comparable. Publishing one page per listing would multiply this site by roughly six while adding nothing a reader can act on — the same model, the same weights, a different host and a different margin. Grouping listings instead produces a page that states the spread and an index that stays readable, and it keeps the numbers on every page verifiable against the methodology.

Need the raw rows instead of the curated view? The full dataset is published as JSON on the API, mirrored as an open dataset, and served to agents through llms.txt.