coding LLM API, code generation model, AI coding assistant
These are the models worth pointing an agent at. Two properties matter more than benchmark position in daily use: reliable tool-calling (agents fail loudly when a model emits malformed calls) and a large output ceiling, because a patch is usually long. All models here work with Claude Code, Cursor, Cline and similar clients by changing only base_url — see the integrations section for per-client setup.
Prices are per 1M tokens in USD · ¥6.6 = $1