llamaperf

DeepSeek V4.1 Flash 552B (16B active)

on NVIDIA RTX 5070 Ti · Strata · 205,000 ctx

NVIDIA hardware
Oct 30, 2026
Throughput
28.8 t/s gen
Quant
1.75 bpw
KV cache
INT4
System RAM
128 GB

Use cases

vision

Summary

User reports DeepSeek V4.1 Flash at 28.8 t/s average on an RTX 5070 Ti with a Ryzen 9950X and 128GB RAM. Setup is Strata with a 1.75 bpw quant and INT4 KV cache, 8,192 token thinking budget and 205k context; the GPU ran +3000MHz memory, +300MHz core and an 83% power limit. A 16k thinking budget pushed the model onto the NVMe drive instead of RAM and dropped speed to 19 t/s. Gemma 4 E4B runs alongside for vision without affecting DeepSeek's performance.

Compare this run

Medians of plain runs: one device, one request, no speculative decoding, the whole model in the device's memory. The report above is one setup, so read it beside them and not as one of them.

DeepSeek V4.1 Flash 552B · 16B active at 3-bit by hardware

Bars are medians on one scale, 0 to 73.4 tokens/s. Press ▶ to watch text arrive at a row's speed.

Models most run on the NVIDIA RTX 5070 Ti

Bars are medians on one scale, 0 to 137.7 tokens/s. Press ▶ to watch text arrive at a row's speed.

Everything reported on the NVIDIA RTX 5070 Ti →Every DeepSeek V4.1 Flash report →Estimate DeepSeek V4.1 Flash on this card →