Tech2Wild
@Tech2Wild
๐ฎ Tech, gaming, AI, and everything in between. ๐ค Building with it, not just talking about it. ๐ฅ From the mind of @ToNYD2WiLD
๊ฐ์
March 2026
222 ํ๋ก์ ์ค 4.8K ํฌ
Ran TensorFold (EXL3) against my own vLLM TP2 recipe, same prompts:
โก Code 82 vs 69 tok/s
โก JSON 84 vs 71
โก Prose 44 vs 35
๐ Mixed load at C4: 108 vs 72 tok/s (+51%)
๐ Prefill 1,781 vs 1,225 tok/s
๐ง 500K context vs 262K
Two boxes on a desk ๐คฏ
๋ ๋ณด๊ธฐ
