AI models
13 models behind the tools in this directory, grouped by what they handle.
This page lists 13 AI models, the engines that the tools in this directory are built on. By what they handle: text (4), multimodal (3), image (2), embedding (2), audio (1) and video (1). Published by DeepSeek, Black Forest Labs, Anthropic, Google DeepMind and OpenAI and others.
5 of the 13 carry an open licence, which means you can run them on your own hardware and inspect the weights. The rest are reached through an API, so the cost and the rate limits are set by whoever publishes them. The largest context window listed is 2,000,000 tokens, on Gemini 2.5 Pro.
Each model page records its publisher, modality, parameter count where it is public, licence and release date. Where a figure is not published, the page leaves it out rather than guessing.
Text models (4)
DeepSeek-V3
DeepSeek
Open mixture-of-experts model that competes with frontier closed models at low cost.
5.4M downloads
Llama 3.3 70B
Meta
Open-weights LLM, instruction-tuned. Excellent base for fine-tuning.
12M downloads
Qwen 2.5 72B
Alibaba
Open-weights multilingual model from Alibaba's Qwen family.
8M downloads
Mistral Large
Mistral AI
Frontier model from Mistral with strong code and multilingual performance.
Multimodal models (3)
Claude Opus 4.7
Anthropic
Anthropic's most capable model. Long-context reasoning, careful writing, top coding ability.
Gemini 2.5 Pro
Google DeepMind
Google's flagship multimodal model with strong reasoning and 2M-token context.
GPT-4o
OpenAI
OpenAI's flagship omni-modal model with strong text, vision, and audio capabilities.
Image models (2)
Embedding models (2)
Audio models (1)
Video models (1)
Frequently asked questions
- What is the difference between a model and an AI tool?
- A model is the engine. A tool is the product built around one: the interface, the workflow and the billing. Several tools in this directory run on the same model.
- What does an open licence actually allow?
- 5 of these models publish their weights under a licence that lets you download and run them yourself. The exact terms differ by model, including whether commercial use is allowed, so each model page records its own licence.
- What is a context window?
- The amount of text a model can consider at once, counted in tokens. A larger window lets you hand it a longer document or a longer conversation before it starts losing the earlier parts.
- Why is a parameter count missing for some models?
- Several publishers do not disclose it. Where the figure is not public, the page leaves it blank rather than printing an estimate.