deepseek-v4.1POST
DeepSeek V4.1 Flash EXL3 3.75bpw on 3 DGX Sparks: 106 tok/s code
A public 3.75bpw EXL3 build of DeepSeek V4.1 Flash 1017 hits 106 tok/s on code and 73 on prose across three DGX Sparks with TensorFold TP3 at 2,000 MHz clocks; the 2.9bpw two-Spark build did 88/61.
Thomas Eastham
@easthambuildsArchived 9 days agoOriginally posted 2026.10.23
Thomas Eastham @easthambuildsOriginally posted 2026.10.23
@JamesTervit @antirez Nice matchup! If you want a 3-Spark column: my public 3.75 bpw EXL3 build on TensorFold TP3 runs 106 code / 73 prose tok/s in rigmark (clocks capped 2,000 MHz). Our 2.9 bpw did 88 / 61 on 2 Sparks. 🚀 Recipe: https://github.com/easthambuilds/DeepSeek-V4.1-Flash-1017-EXL3-3.75bpw-3x-DGX-Sparks-TensorFold
x.com/easthambuilds/status/2113464078500778208Archived 2026.10.23 12:06