Image: reddit.com · I’m using Claude Code to build an AI accelerator. It now runs Qwen3.5 on an FPGA.
openTPU: an open-source AI accelerator running models on an FPGA
An open-source accelerator built with AI coding agents including Claude Code, now running Qwen3.5 on a Kintex-7 FPGA card.
Credits
Earliest dated source:
The full story
openTPU is an open-source AI accelerator the poster started to learn about accelerator design, developed with AI coding agents including Claude Code. It already runs Qwen3.5 and other open-weight models on a Kintex-7 FPGA card.
One agent reads the hardware code, timing reports and notes from earlier attempts and proposes a change, another implements it, and automated checks compare the hardware with a Python simulator and measure performance. The poster still has to nudge larger architectural changes, and verification had blind spots: floating-point bugs that the inference tests missed. The repo holds the hardware, instruction set, simulator, compiler, profiler and host tools, and can be run in the simulator without an FPGA.
How it was done
Tools
Themes
Loading themes…
Sources
From a short post by the maker, so the details are as they tell them.
- Reddit: I’m using Claude Code to build an AI accelerator. It now runs Qwen3.5 on an FPGA.Reddit poster: ScarionnSPosted
The maker's post on r/ClaudeAI.
If it disappears: archived copy of Reddit: I’m using Claude Code to build an AI accelerator. It now runs Qwen3.5 on an FPGA. (opens in a new tab)
- The code: github.comRepository account: FeSens
Linked in the post.