Phoenix Grove API
Private by default. US-based model hosting and data storage.
Endless messages for your coding tools. No message counting, on any plan. Everything runs at full speed: big 5-hour windows, a weekly allowance that resets on the same day every week, and unused usage banks up to ten days. Up to 3.7B tokens a month for coding use on Flash models (like DeepSeek V4.1 Flash 1017) on the Pro plan, and 165M+ on advanced models (like MiMo V2.6 Pro). Or pay per token: the simple way in, $5 minimum, no subscription. The plan stretches further: up to 30% more usage than the same spend per-token. Your slow weeks feed your busy ones. For heavy, regular users, the coding plan is often the better value.
23 open models spanning chat, vision, text to speech and embeddings. Change two lines (base URL + key) and your existing OpenAI SDK code just works. Claude Code, Codex and Ollama-compatible apps get native endpoints of their own.
Base URL: https://api.pgsgrove.com/v1
Auth: Authorization: Bearer pgsk_your_key
Flagship models
mimo-v2.6-pro(Xiaomi): The newest trillion-parameter flagship: deep reasoning, vision, and a million-token window, fresh off the September release. From $0.56/1M input.deepseek-v4.1-flash-1017(DeepSeek): DeepSeek's October 17 update to V4.1 Flash: leaner thinking, quicker finishes, images and tools, a million-token window, at about half the price of V4.1 Flash. From $0.13/1M input.mimo-v2.6-flash(Xiaomi): One of the quickest models on the roster: near-instant responses with a full million-token window, at a price that invites experimentation. From $0.18/1M input.glm-5.2(Z.ai): Million-token context, strong tool use, priced well under the maker’s own rate. From $0.65/1M input.glm-5.3(Z.ai): Z.ai's newest full-size release: the flagship class pushed further on coding, security work, and long agentic runs, with the same million-token context. From $1.6/1M input.deepseek-v4.1-flash(DeepSeek): DeepSeek's brand-new V4.1 Flash: the causal encoder-decoder release everyone is benchmarking, with a million-token window and a near-free cache read that rewards agentic loops. From $0.26/1M input.
Coding plan · All 23 models & pricing · Quickstart · Native protocols · Full documentation