DeepSeek-V4-Flash Ported to Run on AMD Strix Halo APU
Fit-Produce420 · reddit · 2026-07-31
A developer attempted to run a quantized DeepSeek-V4-Flash-GGUF model on AMD's latest Strix Halo hardware.
According to the feedback, the model can support a context length of approximately 64K on a single Strix Halo, achieving an inference speed of about 35 tokens/s. While not exceptionally fast, this local deployment setup is considered practically usable if the model's capabilities hold up.
More from Infra
- Same 1,000 Cloudflare Durable Objects: $4,050/month cut to $22 with one trick — threepointone · 2026-10-16
- NVIDIA ships FP4-quantized Qwen3.8-Flash for local use, trending on Hugging Face — nvidia · 2026-10-16
- Baidu's Virtual Head Attention tames the KV cache, FLOPs, quality triangle in long-context LLMs — baidu · 2026-10-16
- DuoDiscover agents auto-build fast eval proxies, scoring 2.68x TPU kernel speedup — berkeley_ai · 2026-10-16
- Google Raises 2026 Capital Spending to Up to $205 Billion — emmanuelvivier · 2026-10-16
- Google open-sources AX, a Kubernetes-style orchestrator for billions of AI agent tasks — thisdudelikesAI · 2026-10-16