Short answer

The directory currently tracks 51 models from 13 providers. 48 have verified first-party API prices and 51 have an available OpenRouter route. Every row links to a crawlable model page with separate channel data, verification dates, context, lifecycle notes, and related models.

Browse 51 tracked models

Search by model, provider, or model ID. Use provider filters to narrow the list without leaving the page.

Showing all 51 models

AC

Alibaba Cloud

5 tracked models

Provider pricing →
A

Anthropic

8 tracked models

Provider pricing →
C

Cohere

2 tracked models

Provider pricing →
D

DeepSeek

2 tracked models

Provider pricing →
G

Google

8 tracked models

Provider pricing →
M

Meta

3 tracked models

Provider pricing →
M

MiniMax

2 tracked models

Provider pricing →
MA

Mistral AI

4 tracked models

Provider pricing →
MA

Moonshot AI

1 tracked model

Provider pricing →
O

OpenAI

8 tracked models

Provider pricing →
TA

TypeSafe AI

1 tracked model

Provider pricing →
Z

Z.ai

1 tracked model

Provider pricing →
X

xAI

6 tracked models

Provider pricing →

How should you use the model directory?

Start with a model family or provider, then open the detail pages that match your budget and context requirements. The listed prices provide a consistent first pass, but they do not make different models equivalent. Test quality, latency, tools, regional availability, and data policies before moving production traffic.

New model releases can enter this directory without creating a separate short-lived site. Their detail pages capture launch demand, while the directory, provider pages, calculator, and price history retain value when the next model generation arrives.

AI model directory questions

What information is included for each model?

Each model page includes official and OpenRouter token rates when available, cached-input pricing, context, lifecycle and pricing status, source links, verification dates, an example workload, and related provider models.

Does the directory rank models by quality?

No. It organizes verifiable pricing and catalog metadata. Capability, latency, reliability, tools, safety behavior, and output quality must still be tested on the intended workload.

Why are some official prices unavailable?

Some open-weight models do not have one universal first-party hosted API. The directory labels those records instead of presenting an OpenRouter rate as an official provider price.

How often is the model directory updated?

The scheduled workflow checks official sources and OpenRouter daily. Source failures preserve the last verified value, and suspicious changes wait for review.