Skip to content

My neuronets Registry Spaces Docs FAQ

Sign in

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

ProductCreate a neuronet Model catalogue Datasets Spaces My payouts

ResourcesDocumentation API Timings & statuses FAQ

Networkaetron.io Privacy & verification

© 2026 AETRONTestnet

Models/gemma-4-31b-w4a16

G

Gemma 4 31B (official QAT W4A16)

Google · Gemma 4

Create a neuronet with this model →

LLMGemma-TermspublicTrusted31B params

chat flagship int4 quantized

Ready-made official QAT W4A16 build of Gemma 4 31B.

Ready-made official QAT W4A16 build of Gemma 4 31B.

At a glance

Task: LLM (InferenceV1)
Params: 31B
License: Gemma-Terms (commercial use OK)
Hardware: 26 GB+ VRAM
dtype: Int4

Create a neuronet with this model →

Overview

Task: LLM · InferenceV1
Family: Gemma 4
Tier: Trusted
Status: active

Runtime spec

dtype: Int4
attention: FlashAttn2
compile: ReduceOverhead
seed: 0
gpu_arch: Sm89
Runs on: Consumer, Datacenter

Hardware

Min VRAM: 26 GB
Recommended: 29 GB
Tensor parallel: ≥1
TEE archs: Sm89, Sm90

Verification defaults

Tier: B
Trust: Level1
Security: Standard
Privacy: P0
Min pulse score: 100

License

spdx: Gemma-Terms ↗
Commercial use: ✓
Redistribution: ✓

Source & provenance

Repo: google/gemma-4-31B-it-qat-w4a16-ct ↗
Revision: 52f3f65bc7
Added: 2026-08-03
Verified: 2026-08-03