---
tags:
- blackhole
- p300x2
- tt-model-cache
- tt-model-container
- vllm-plugin
---

# qwen3-32b-p300x2

Qwen3-32B is a dense instruction and reasoning model for general chat, coding, and tool use, served on a four-chip Tenstorrent Blackhole mesh.

Runs on **p300x2** (mesh `P300x2`) — 131,072-token context, up to 32 concurrent sequences.

Packaged and published with [tt-model-manager](https://github.com/tenstorrent/tt-model-manager) 0.1.0 (manifest schema 5.1).

## Quickstart

```bash
tt-model pull  stisiTT/qwen3-32b-p300x2 --with-weights
tt-model serve stisiTT/qwen3-32b-p300x2
```

`pull --with-weights` downloads the Docker image and the [`Qwen/Qwen3-32B`](https://huggingface.co/Qwen/Qwen3-32B) weights at `9216db5781bf21249d130ec9da846c4624c16137` (into your HF cache; they are not in the image). `serve` starts an OpenAI-compatible server on port 20000 (or the next free port, if that one is busy); the first start compiles kernels for your device, which takes several minutes, and the server is ready when it logs `Application startup complete`.

## Provenance

The exact sources the image was built from — `code/` in this repo is byte-identical to the model code inside the image:

| component | built from |
| --- | --- |
| tt-metal | [`7d54a363f0cb5ccb0fc0ff8a63fb35398a4ba0b4`](https://github.com/tenstorrent/tt-metal/commit/7d54a363f0cb5ccb0fc0ff8a63fb35398a4ba0b4) *(dirty tree — the image includes uncommitted changes)* |
| vLLM | [`v0.26.0`](https://github.com/vllm-project/vllm/releases/tag/v0.26.0) |
| vllm-tt-plugin | a local checkout — commit not published |
| `code/` digest | `2c329baf1a4b0bea` (sha256, first 16 hex digits) |
| built | 2026-09-10T01:37:02+00:00 by tt-model 0.1.0 |