Compare AI models by cost, context window, modality, and capabilities

Add two or more models, then read each row across the comparison table. The page can handle larger comparison sets with horizontal scrolling, but two to six models usually produce the clearest decision. Use a larger set for initial screening and a smaller set for the final choice. If you came from the directory, saved models can load automatically as your starting comparison list.

The comparison table highlights the lowest known input cost and output cost among the selected models in green. Unknown cost values are ignored, so a missing price is never treated as the cheapest option. Input and output costs are evaluated separately because the cheapest model for prompts may not be the cheapest model for generated responses. Use the highlighted cells as a cost advantage signal, then check whether the model still fits your modality and limit requirements.

Lowest input cost means the model has the cheapest known cost for processing prompt or input tokens among the models you selected. This matters most for retrieval, long documents, agent loops, and workloads that send a lot of context into the model. A low input cost does not guarantee low total cost if the model produces expensive or very long outputs. Compare input cost with context window and output cost before choosing.

Lowest output cost means the model has the cheapest known cost for generated tokens or outputs among the selected models. This matters most for chatbots, content generation, code generation, summaries, and any workflow where responses are long. A model can have a low output cost but still be a poor fit if it lacks the required modality, context window, or tool-calling support. Treat output cost as one part of the final tradeoff.

After cost, compare modality fit, context window, max output, reasoning support, tool calling, open-weights status, knowledge cutoff, release date, and last updated date. Modality fit is the first blocker because a cheap model is not useful if it cannot accept your input or produce your required output. Context window and max output determine whether the model can handle long prompts or long responses. Capability flags help identify models that fit agents, tool workflows, local deployment, or open-weight requirements.

Your comparison list is stored locally in your browser. It lets models added from the directory appear automatically on this page when you use the same device and browser. The saved list is only a convenience layer for comparison setup, not an account-level setting. If you clear browser storage or switch browsers, you may need to add the models again.