Python SDK & Hosted API now available for Alpie Core (4-bit, 32B)
We are so happy to announce we have added a hosted API, official Python SDK, and CLI for Alpie Core to make it easier to use the model in real systems, beyond local inference and notebooks.
Alpie Core is a 32B reasoning model trained and served entirely at 4-bit precision. Instead of post-hoc compression, it’s optimised end-to-end for low-precision reasoning, which allows it to retain strong multi-step reasoning while using far less memory and infrastructure.
What’s new:
- Official Python SDK (sync, async, streaming)
- OpenAI-compatible hosted API
- CLI for quick testing and automation
- 65K context support
The model is open source and still available via Hugging Face and Ollama for local runs. The hosted API is intended for production-style usage and evaluation.
To make testing easier, the first API key includes 5 million free tokens so people can run longer workflows and not just sample prompts.
We’d especially appreciate feedback on:
- long-context reasoning behaviour
- agent and tool-using workflows
- failure modes and stability under load
- SDK ergonomics
Links:
- API access: https://169pi.ai
- SDK access: https://github.com/169Pi/Pi169-SDK
- Playground (quick testing): https://playground.169pi.ai
We are actively iterating based on real-world usage, so concrete feedback or failure cases are extremely helpful.
I want to contact you regarding it, can we have a voice call or a personal chat to talk about this? @Chirag2207