models · google

gemma-4-31B-it

image-text-to-text model harvested from Hugging Face (trending). Catalogued at Quarantine — unverified, pending attestation.

Compare Verify

What this registry has checked

Quarantine. We hold the publisher’s own published hash for these weights and have recorded it against this entry. We have not downloaded and hashed the weights ourselves, so the hash is the publisher’s claim, not our measurement.

Method	hf-published-sha256

Re-check this yourself → Source → Full record →

What is recorded

Publisher	google
Origin	United States
Licence	apache-2.0
Parameters	31,273,088,876 (31.3B)
Context window	262,144 tokens
Modality	text+image+video->text
Native precision	BF16
Pulls recorded here	9,794,803
Downloads reported upstream	8,818,960

Measured on real hardware

None of these runs are ours. Each figure is the reporter’s, linked to their own evidence, under their own trust label. The number shown is single-stream decode — one user, concurrency 1 — because aggregate throughput across many users is a different quantity and reads as far faster than anyone will see.

Reported to local-ai-registry

Hardware	Engine	Build measured	Decode, 1 stream	Label	Reported by
GeForce RTX 5090	llama.cpp	UD-Q4_K_XL-QAT	68.5 tok/s @ 8,192	candidate	localmaxxing
GeForce RTX 5090	llama.cpp	Q6_K_XL	51.7 tok/s @ 2,048	candidate	localmaxxing
GeForce RTX 5090	llama.cpp	UD-Q6_K_XL	46.5 tok/s @ 8,192	candidate	localmaxxing
GeForce RTX 3090	llama.cpp	GGUF Q4_K_M	38.9 tok/s @ 2,048	candidate	localmaxxing
GeForce RTX 5090	llama.cpp	Q8_0	38.0 tok/s @ 2,048	candidate	localmaxxing
Radeon AI PRO R9700	llama.cpp	Q4_K_M	28.8 tok/s @ 2,048	candidate	localmaxxing
Apple M3 Ultra 96GB 60-core GPU	mlx	MLX 4bit	27.6 tok/s @ 8,256	candidate	exo-postgres
Radeon AI PRO R9700	llama.cpp	Q4_K_XL	27.6 tok/s @ 2,048	candidate	localmaxxing
Apple M3 Ultra 96GB 80-core GPU	mlx	MLX 4bit	27.1 tok/s @ 8,256	candidate	exo-postgres
Apple M5 Max 128GB	llama.cpp	UD-Q4_K_XL-QAT	26.3 tok/s @ 2,048	candidate	localmaxxing

From local-ai-registry (MIT), commit 124959d: 57 runs across 16 machines, 0 validated — model revision and runtime pinned, launch accepted — and 57 candidate, their label for useful evidence without a reproducible-launch promise. 38 more not listed.

Measured by local.ai

Hardware	Build measured	Decode @ 8K	Memory @ 8K	Decode mode
NVIDIA RTX PRO 6000 Blackwell 96GB	Q4_K_M	62.8 tok/s	20.8 GB	ordinary
NVIDIA GeForce RTX 5090 32GB	AWQ-4bit	61.9 tok/s	28.2 GB	ordinary
NVIDIA RTX PRO 6000 Blackwell 96GB	AWQ-4bit	59.6 tok/s	89.6 GB	ordinary
NVIDIA RTX 6000 Ada 48GB	AWQ-4bit	38.4 tok/s	43.2 GB	ordinary
NVIDIA RTX 6000 Ada 48GB	Q4_K_M	37.9 tok/s	39.3 GB	ordinary
NVIDIA RTX PRO 6000 Blackwell 96GB	FP8	37.1 tok/s	90.9 GB	ordinary
Mac Studio M3 Ultra 96GB · 60-core GPU	Q4_K_M	26.7 tok/s	32.6 GB	ordinary
Mac Studio M3 Ultra 96GB · 80-core GPU	Q4_K_M	26.5 tok/s	47.0 GB	ordinary

Measured by local.ai, not by us: 67 runs, 59 more not listed. Each names its engine pinned by image digest and the exact serve command, in the full record.

Elsewhere in the registry

All models → Browse with filters → What will run on my hardware? →

Registry record as of 2026-09-01. Identity, licence and parameter count are properties of a release and do not move; tier can, and the verify link above re-checks it against the live registry rather than this page.