DeepSeek V4.1-Flash hits 49 tok/s decode on dual Radeon R9700 in local test
Longjumping-Elk-7756 · reddit · 2026-10-26
A Reddit user benchmarked DeepSeek V4.1-Flash (IQ2XXS, 2bpw) locally on 2× Radeon AI PRO R9700 (32GB each, 64GB total VRAM) with TP=2 and 128GB RAM.
Key numbers:
- Decode: 28.4 tok/s at 1K context, 31.3 at 4K, 49.1 at 8K (with DSpark on)
- Prefill: 137.6 tok/s at 1K, 265.7 at 4K, 344.3 at 8K
- DSpark boosts decode by +35%/+44%/+52% at 1K/4K/8K, though 8K prefill dips -4%
- Stack: Strata official runtime, median of 3 runs
A rare public data point for running the latest DeepSeek quantized on high-end consumer AMD cards.
More from Infra
- Running Qwen3.8B across 3 GPUs and 130GB RAM only hits 18 T/s — MageLD · 2026-10-26
- Used RTX 3090s Now Top $1500 on eBay as Owner Hunts a Replacement — like_a_tensor · 2026-10-26
- Ternary 1.58-bit Nemotron 30B-A3B GGUF Runs 368K Context on 12GB GPU — pmttyji · 2026-10-26
- Anthropic built Scion to replace git as AI-driven merges jumped from 2,500 to 30,000/day — bibryam · 2026-10-26
- NVIDIA shows RTX Spark powering sim-to-real robotics: Microduck robot learns somersault in simulation — Aiden_Tech_Ai · 2026-10-26
- MiniMax H3 community adds 8-step reference control, 4-step distill, Apple Silicon support — Hailuo_AI · 2026-10-26