Open in tt-awesome →

Wolfpine

community★ featured
by mesham · Python · BSD-3-Clause · 14⭐ ·

A pure-Python framework for building cycle-approximate simulators of AI accelerators, with Wormhole and Blackhole backends written from the public Tenstorrent ISA documentation. Runs real tt-metal programs unmodified on the full worker grid (72 cores on Wormhole, 130 on Blackhole) with no hardware, and because it is Python throughout, a driver script can inspect simulator state directly. The `wolfpine run` CLI switches a tt-metal build between simulated Wormhole, simulated Blackhole and real silicon. Not cycle accurate by design. Formerly tt-sim; renamed to Wolfpine in 2026 and unrelated to Tenstorrent's official `ttsim`.

📦 Repo
LATEST v2.2 2026-10-18T10:26:16Z Release notes ↗
3 previous releases
v2.1 2026-09-24T12:28:22Z
v2.0 2026-08-18T14:07:12Z
v1.0 2026-05-11T13:07:42Z
See all releases on GitHub ↗
simulator architecture no-hardware research python
wormhole blackhole