DeepSeek-V4.1-Flash-1017 GGUF quants appear on Hugging Face with 473GiB Q8_0
realmrfakename · x · 2026-10-18
mrfakename uploaded GGUF quantizations of DeepSeek-V4.1-Flash-1017 to Hugging Face, implying a new DeepSeek model (V4.1-Flash, build 1017) runnable locally via llama.cpp. Tiers include:
- Q80 (473 GiB, 10 files): everything Q80 except routed experts kept at native MXFP4, closest to original weights
- Q4KM (414 GiB): routed experts already 4-bit, so savings are limited
- Q3KM (323 GiB)
Tagged deepseek architecture, Image-Text-to-Text, conversational, MIT license. The leaked chat template shows <think> reasoning mode, configurable reasoning effort, and DSML-format tool calls. Unverified as an official release; community quants.
Related event: GGUF Quantizations of DeepSeek-V4.1-Flash Leak onto Hugging Face(2 posts)→
More from Infra
- China Rolls Out 'Token Factories' to Power Its AI Economy — nordicinst · 2026-10-18
- Running 20 concurrent low-latency regional-language voice agents on a single L4 GPU — DistrictUnique5259 · 2026-10-18
- AgentCave puts your scattered Macs and GPU boxes in one virtual room — noahsolomon · 2026-10-18
- Real AI demand estimated at ~$189B in 2026, over half from OpenAI and Anthropic — builderjaydub · 2026-10-18
- Twinax cable shipments in AI backplanes hit 85,200 km/year at 224G — moon-length in 4.5 years — jwt0625 · 2026-10-18
- DeepSeek-4.1-Flash Runs on 4x RTX Pro at 300+ TPS With NVMe Offload — TheZachMueller · 2026-10-18