๐๏ธ "Medium-Sized" LLM Burners Coming Soon! ๐ฅ
This Could Make Local HyperToken Generation a Reality. โก๏ธ NVIDIAโs worst nightmare? ๐ฑ
โ๏ธ Application-Specific Hardware
Taalas new PCIe ASIC board would burn the entire medium-sized Qwen 3.5-27B LLM straight into silicon ๐คฏ (already doing it with small models)
Taalos said medium models on ASIC would be available in their lab by Spring '26.
๏ฟฝ๏ฟฝ๏ฟฝ๏ฟฝImagine:
๐ซ No more loading weights
๐ ~10,000 Tokens Per Second locally (Llama 3.1 8B already @ 17,000 tps)
๐ป Standard PC slot, ultra-low power (10x less) ๐
๐ 100% offline with no cloud, no GPU farm
๐ฐ Reddit unit cost rumor $300 to $400
๐ฅ๏ธ Imagine HyperToken generation on your desktop.
๐ค AI agents that think at light speed. โก๏ธ Are you ready? ๐
์ด์ฏค๋๋ฉด ์ต์๋์ธ @obsdmd ์์ ํ๊ตญ๊ด๋ จ ๊ฐ์ขํ๋์ฏค์ ๋ด์ค์ผ์๊ฐํฉ๋๋ค
At this point, I think Obsidian should release at least one course related to Korea.