AIHacker Newsabout 3 hours ago
Open-source AI builds own inference chip
7 min read
An open-source project named openTPU successfully developed an AI accelerator capable of running its own inference. The team tested the design on a Xilinx Kintex-7 FPGA card, achieving decode speeds up to 11.07 tokens per second for LFM2.5-8B-A1B models.
Level
Hype check
Tap to vote and see what everyone thinks.
