Skip to main content
Back to trending

openTPU

An open-source AI accelerator, developed by AI: RTL, ISA, simulator, compiler and profiler in one repo. Runs Qwen3, LFM2.5 and Qwen3.5 on a Kintex-7 PCIe card.

View on GitHub
Stars
451
Forks
25
+ today
+1
Created
16d

Ranking data as of October 9, 2026 (UTC).

Star History

Today, hour by hour

451 stars
450451

Overview

An open-source AI accelerator, developed by AI: RTL, ISA, simulator, compiler and profiler in one repo. Runs Qwen3, LFM2.5 and Qwen3.5 on a Kintex-7 PCIe card. It ranks #5440 on GitTiger, gaining +1 star on October 9, 2026 (UTC).

The project is written in Verilog and has 25 forks. It was created 16d ago.

AI-generated summary of the repository README — may contain errors.

What it is

`FeSens/openTPU` is described as an open-source AI accelerator that brings hardware and software components together in one repository. The description lists RTL, an instruction-set architecture, a simulator, a compiler, and a profiler.

Capabilities described

The project description says it runs Qwen3, LFM2.5, and Qwen3.5 on a Kintex-7 PCIe card. This presents a specific combination of accelerator hardware and model support, but the supplied entry does not include independent validation, throughput figures, accuracy results, system requirements, or a complete model-compatibility list. Tags and partial preview text do not supply enough detail to extend those claims. Prospective users should consult the repository documentation for supported configurations and current implementation status.

Getting started

No setup or build instructions are included in the entry. See the repository README for hardware prerequisites, firmware or bitstream guidance, software dependencies, and steps to run the listed models; none should be inferred from the brief description.

Installation
git clone https://github.com/FeSens/openTPU.git
cd openTPU
# see README for setup

Frequently Asked Questions

Which components are included, according to the description?

RTL, ISA, simulator, compiler, and profiler.

Which models are named as running on the accelerator?

Qwen3, LFM2.5, and Qwen3.5.

What hardware is specified?

A Kintex-7 PCIe card; further hardware and system requirements were not supplied.