Every model below is served through the same OpenAI-compatible endpoint with one API key. Prices are in USD per 1M tokens; the console shows the live price applied to each request. The catalog grows as new providers are onboarded.
Prices as currently published; the live per-request price in the console prevails. Where a model's route supports prompt caching, cached input tokens bill at the cache-read price shown for each model in the table, never above the provider's own cache-read price except for DeepSeek V4 Pro, DeepSeek V4.1 Flash and Kimi K3, and cache writes carry no separate charge.
Tags next to a model code (streaming, tools, JSON schema) mean that capability passed automated checks on every route currently serving the model through the chat completions endpoint, within the last seven days. A model without a tag has not been verified yet; that does not mean the capability is missing.
Context caps: models marked ≤120K / ≤240K accept inputs up to that size, at a single flat rate matching Alibaba Cloud's official price for the corresponding usage tier. Larger inputs return a context_length_exceeded error before any charge. For 1M-context work, qwen3.7-max carries the official single rate with no cap.
Models with a details link have their own page — price, a ready-to-run call, and the closest alternatives. Those pages are rolling out model by model; every model in the table is callable today either way.
