Ways to run a model

Choose models and see the three ways to pay for them: the provider's API, a gateway, or a subscription plan. They are billed differently, so they sit side by side, not in one ranking.

Start with a ready-made set

Everyday chatClaude Sonnet 5.5 · GPT-6.1 Sol · Gemini 3.8 FlashThe usual suspects, across all three ways to pay.Open weights firstGLM-5.3 · GLM-5.3-Flash · Kimi K2.6 · Qwen3 Coder Next · Llama 4 Maverick 17b 128e InstructWho resells five open-weight chat models, and which plans include them.Coding side by sideKimi K2.7 Code HighSpeed · Qwen3 Coder Next · GPT-6.1 Sol · Claude Sonnet 5.5Coding-tuned models as APIs, gateways and flat-fee coding plans.Small and fastClaude Haiku 5.5 · Gemini 3.5 Flash Lite · GLM-5.3-FlashThe latency tier, where every dollar per million counts.

Add a model (up to 6)

Loading…

These are three different billing models. API prices are per token and grow with use. A plan is a flat fee for an allowance the provider defines, in messages, prompts or hours rather than tokens, and a chat plan usually does not include API access. We do not convert one into the other; the break-even page compares them by monthly spend, as an estimate.