How often is this AI model list updated?

New models, version bumps, pricing changes, and benchmark results are added as soon as they are published — typically within hours of an official announcement. Once the data is in, the page refreshes within 5 minutes so visitors always see the latest information.

What do the "Open Source" filter options mean?

A model can be open source (weights publicly available) yet still restrict commercial use through its license — for example, some Llama variants prohibit large-scale commercial deployment. The three options reflect this: "Free for commercial use" means no restrictions for production; "Paid commercial" means a license fee applies; "Not for commercial use" means the model cannot legally be used in commercial products.

How do I pick the right model for my use case?

Start by filtering on capability (chat, coding, reasoning, multimodal) to narrow the field, then compare the top candidates on the benchmark closest to your workload. For production, also weigh API pricing, context length, and license — a model that wins on benchmarks may still lose on total cost of ownership.

Which organizations' models are included?

The list covers mainstream model publishers worldwide — including OpenAI, Anthropic, Google, Meta, Mistral, DeepSeek, Alibaba (Qwen), Zhipu (GLM), Moonshot (Kimi), and many others. Use the publisher filter to narrow to a specific lab.

Can I see the full benchmark results for a specific model?

Yes — click any model card to open its detail page, where you will find the complete benchmark table, parameter sizes, context window, license, API pricing, and links to the official paper, Hugging Face, and GitHub repository.