Skip to content
AETRONConstructor
My neuronetsRegistrySpacesDocsFAQ
Sign in
AETRON

A constructor for AI services on AETRON’s decentralised network of GPU miners.

ProductCreate a neuronetModel catalogueDatasetsSpacesMy payouts
ResourcesDocumentationAPITimings & statusesFAQ
Networkaetron.ioPrivacy & verification
© 2026 AETRONTestnet
Models/deepseek-v4-flash-nvfp4
D

DeepSeek-V4-Flash (NVFP4-FP8 (compressed-tensors))

DeepSeek · DeepSeek
Create a neuronet with this model →
LLMMITpublicTrusted— params
fastmoenvfp4quantized

Ready-made NVFP4-FP8 (compressed-tensors) build - runs on the fleet stack (compressed-tensors). Blackwell window.

Ready-made NVFP4-FP8 (compressed-tensors) build - runs on the fleet stack (compressed-tensors). Blackwell window.

At a glance

  • Task: LLM (InferenceV1)
  • Params: —
  • License: MIT (commercial use OK)
  • Hardware: 181 GB+ VRAM
  • dtype: Nvfp4
Create a neuronet with this model →
Overview
Task
LLM · InferenceV1
Family
DeepSeek
Tier
Trusted
Status
active
Runtime spec
dtype
Nvfp4
attention
FlashAttn2
compile
ReduceOverhead
seed
0
gpu_arch
Sm120
Runs on
Datacenter
Hardware
Min VRAM
181 GB
Recommended
188 GB
Tensor parallel
≥4
TEE archs
Sm120
Verification defaults
Tier
B
Trust
Level1
Security
Standard
Privacy
P0
Min pulse score
100
License
spdx
MIT ↗
Commercial use
✓
Redistribution
✓
Source & provenance
Repo
RedHatAI/DeepSeek-V4-Flash-NVFP4-FP8 ↗
Revision
39087b5c25
Added
2026-08-03
Verified
2026-08-03