deepseek-v4.1POST

DeepSeek V4.1 Flash on 4x DGX Spark: 97 tok/s and a 29.4M-token KV pool

Deployed with the TensorFold v0.3.5 recipe and 4.0bpw pack on four DGX Sparks, hitting 97 tok/s for one user, 231 tok/s at 16 users, and a 93K prompt read in 37s.

Tech2Wild

@Tech2WildArchived 2 days agoOriginally posted 2026.10.27

Tech2Wild @Tech2WildOriginally posted 2026.10.27

Open original on X

DeepSeek V4.1 Flash on 4x DGX Spark, now on @bertholomusai's TensorFold v0.3.5 recipe + his 4.0bpw pack: • 29.4M-token KV pool (was 2M) • 97 tok/s at 1 user, 144 on code, 231 at 16 users • 93K prompt read in 37s, ~2s replies during long reads https://github.com/tonyd2wild/DeepSeek-V4.1-Flash-TP2-and-TP4-DGX-Sparks

x.com/tech2wild/status/2114866340645278008Archived 2026.10.27 12:08