gemma-4-12B

12.0B parameters, several kinds of input and output, 256K tokens of context.

Verified 2026-10-25 Rank #277 in LLMs · score 84.7 method v1 Sources 3 Downloads 30d 69.6k · #356 of 619 tracked

Changed this week: 1 new quantized build on the Hub (50 in all).2026-10-24

Can I run it?7.2GB at INT4Estimatefits a 16 GB cardmethodCommunity: I tested Gemma4 12B on my 8GB GPU, now I don't want to go back to smaller models →

Can I use it?Apache 2.0Permissive; patent grantlicence file2026-10-24

Do people use it?69.6kdownloads, 30 days#356 of 619 we track · below the median+16 likes, downloads −11% this week816 likesHugging Face2026-10-24

Worth reading

Picked by rule from 42 tracked pieces: the lab's own post, the most shared, a field report, the latest.

The lab's own post

Gemma 4 12B: A unified, encoder-free multimodal model

Google releases Gemma 4 12B, an Apache 2.0 licensed model that has crossed 150 million downloads and runs locally on hardware with 16GB of VRAM or unified memory.

blog.google 2026-06-03 1,062 HN pts · 401 comments 584 words

Most discussed · 868 HN pts

Gemma 4 on iPhone

apps.apple.com 2026-04-05 868 HN pts · 234 comments about the gemma 4 family family

Latest · 3 months ago

Show HN: Cactus Hybrid: We taught Gemma 4 to know when it's wrong

Cactus Hybrid introduces on-device models that score every answer with a confidence between 0 and 1 to decide when to hand off to a bigger model.

Weights on the meta device crash the confidence read because probes score generations outside the module forward path.
The probe is a patch you compile into the llama.cpp engine rather than a standalone library component.

github.com 2026-07-22 191 HN pts · 44 comments 714 words about the gemma 4 family family

25 best of 42 tracked

The 3 picks above, plus these 22. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.

CodeShow HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Macgithub.com · 2026-07-29
ArticleRunning Gemma 4 26B at 5 tokens/sec on a 13-year-old Xeon with no GPUneomindlabs.com · 2026-07-15
ArticleGemma 4 QAT models: Optimizing compression for mobile and laptop efficiencyblog.google · 2026-06-05 · 406 HN pts · official
ArticleAccelerating Gemma 4: faster inference with multi-token prediction draftersblog.google · 2026-05-05 · 687 HN pts · official
ArticleGoogle releases Gemma 4 open modelsdeepmind.google · 2026-04-02 · 1,812 HN pts · official
CodeGemma 4 E2B inference in 700 lines of Cgithub.com · 2026-08-27 · 18 HN pts
ArticleFable 5 pushed Gemma 4 to 255 tok/s on WebGPUxcancel.com · 2026-06-18 · 48 HN pts
ArticleGoogle Gemma 4 Runs Natively on iPhone with Full Offline AI Inferencegizmoweek.com · 2026-04-15 · 303 HN pts
CodeApril 2026 TLDR Setup for Ollama and Gemma 4 26B on a Mac minigist.github.com · 2026-04-03
ArticleShow HN: Determinstic LLM inference for lowest price Gemma 4, with Windows XPtokendelivery.ai · 2026-09-12 · 9 HN pts
ArticleI ran Gemma 4 as a local model in Codex CLIblog.danielvaughan.com · 2026-04-12
ArticleRunning Gemma 4 locally with LM Studio's new headless CLI and Claude Codeai.georgeliu.com · 2026-04-05
ArticleGemma 4 on Cerebras - The Fastest Inference Is Now Multimodalcerebras.ai · 2026-06-30
CodeShow HN: Gemma 4 Multimodal Fine-Tuner for Apple Silicongithub.com · 2026-04-07
ArticleShow HN: Prompt-to-Excalidraw demo with Gemma 4 E2B in the browser (3.1GB)teamchong.github.io · 2026-04-19
CodeShow HN: Local privacy-first Microsoft Recall alternative with Gemma 4github.com · 2026-07-04
ArticleGemma 4 12B: The Developer Guidedevelopers.googleblog.com · 2026-06-03 · official
ArticleGemma 4: Byte for byte, the most capable open modelsblog.google · 2026-04-02 · official
ArticleA Visual Guide to Gemma 4 12Bnewsletter.maartengrootendorst.com · 2026-06-03
CodeShow HN: Local LLM code-generation with Gemma 4 e2B via JSON AST to Clojuregithub.com · 2026-05-19
ArticleShow HN: Running Gemma-4 26B at 124 tokens/SEC on a CPU, no GPUapeg.dev · 2026-06-30
ArticleShow HN: I made a Gemma 4 Mac app that names screenshots with local AIsnapname.app · 2026-05-31

Show all 22

Where it's used

Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.

No API provider serves it, so its downloads are people running it themselves: 50 quantized builds and 16 Spaces.

As a service0no API provider listed
On own hardware50quantized builds
Built on10688 fine-tunes · 18 adapters · 0 merges
In apps16Hugging Face Spaces

Top Spaces running it: build-small-hackathon/multi-agent-lab ♥ 6 · build-small-hackathon/chalchitra ♥ 2

Will it fit?

Weights only, an estimate: leave room for the KV cache. Method

Fits a 16 GB card at INT4

About 7.2 GB at INT4, 14 GB at FP8, 29 GB at BF16: weights only, so leave room for the KV cache.

Estimate params × bytes per param × 1.2 · how we estimate

The story so far

Weights, every dated article, and repo commits on one rail.

Most coverage came in the first two weeks

Teal is the weights (2026-05-23), amber triangles are 10 dated articles (each links to it), grey ticks are repo commits.

What people built on it

156 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.

156 derivatives on the Hub, most of them fine-tunes

From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.

GGUF Mungert/gemma-4-12B-it-GGUF for llama.cpp, Ollama, LM Studio 30.8k downloads MLX mlx-community/gemma-4-12B-4bit for Apple silicon 15k downloads FP8 lokeshe09/gemma-4-12b-it-FP8-Dynamic for vLLM, SGLang 1.3k downloads

All variants →

Reference

Everything else, folded: the full tables, every source.

Benchmarks

16 rows

Benchmarks 16 rows

measured third-party leaderboard rowauthor-reported number from the model card or eval files (16)No confidence intervals unless the source gave them; settings differ between rows, so compare within a benchmark, not across. Date = the benchmark's own date when it gives one, else when we read it.

Benchmark	Value	Setting / notes	Reported by	Date
author-reported AIME 2026 · Reason through hard problems	77.5% percent	Gemma 4 12B Unified	model author	2026-10-07
author-reported BBEH · Reason through hard problems	53.0% percent	Gemma 4 12B Unified	model author	2026-10-07
author-reported CoVoST	38.5*	Gemma 4 12B Unified	model author	2026-10-07
author-reported Codeforces · Write code with agents	1659 elo	Gemma 4 12B Unified	model author	2026-10-07
author-reported FLEURS (lower is better)	0.069*	Gemma 4 12B Unified	model author	2026-10-07
author-reported GPQA Diamond · Reason through hard problems	78.8% percent	Gemma 4 12B Unified	model author	2026-10-07
author-reported Humanity's Last Exam · Reason through hard problems	5.2% percent	Gemma 4 12B Unified	model author	2026-10-07
author-reported LiveCodeBench · Write code with agents	72.0% percent	Gemma 4 12B Unified	model author	2026-10-07
author-reported MMLU-Pro · Reason through hard problems	77.2% percent	Gemma 4 12B Unified	model author	2026-10-07
author-reported MMMLU · Work across languages	83.4% percent	Gemma 4 12B Unified	model author	2026-10-07
author-reported MMMU-Pro · Understand images and video	69.1% percent	Gemma 4 12B Unified	model author	2026-10-07
author-reported MRCR · Read long documents	43.4% percent	Gemma 4 12B Unified	model author	2026-10-07
author-reported MathVision · Understand images and video	79.7% percent	Gemma 4 12B Unified	model author	2026-10-07
author-reported MedXPertQA MM	48.7% percent	Gemma 4 12B Unified	model author	2026-10-07
author-reported OmniDocBench · Extract text from documents	0.16 other	Gemma 4 12B Unified	model author	2026-10-07
author-reported tau-bench · Call tools and APIs	69.0% percent	Gemma 4 12B Unified	model author	2026-10-07

What the card says

1 cited sentence

Summary generated, cited

The model uses an encoder-free architecture to process text, image, and audio inputs.^[1]

“projecting raw image patches and audio waveforms directly…”

Generated 2026-10-07 15:06 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.

Use-case evidence

8 use cases

What it is good at 8 use cases

Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.

Facts

21 fields

Listed in LLMs

Facts

gemma-4-12B has 12.0B parameters. (safetensors metadata, observed 2026-10-24)
Context length is 262,144 tokens (256K). (config.json max_position_embeddings, observed 2026-10-24)
License: Apache 2.0 (friction: none). Permissive; patent grant. (license file, observed 2026-10-24)
Created 2026-05-23, last modified 2026-07-15. (HF API, observed 2026-10-24)

Pipeline tag	any-to-any
Library	transformers
Architectures	Gemma4UnifiedForConditionalGeneration
Model type	gemma4_unified
Dtype	bfloat16
Params by dtype	BF16: 11,959,730,224
Size band	L
MoE	no
Context length	262,144 (config.json max_position_embeddings)
Papers	arXiv:2607.02770
License	Apache 2.0 (apache-2.0)
Hub snapshot	2026-10-25

Servability

vLLM · SGLang

Servability

vLLM: supported · Gemma4UnifiedForConditionalGeneration · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Gemma4UnifiedForConditionalGeneration · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed

vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.

Hardware fit

3 estimates

Hardware fit ESTIMATE

Weights-only memory for gemma-4-12B: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)	29 GB · Single high-end GPU (<80 GB)
FP8 (1 byte/param)	14 GB · Single GPU (<24 GB)
INT4 (0.5 bytes/param)	7.2 GB · Single GPU (<24 GB)

Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.

All variants

40 variants

Lineage

Original model Family: google/gemma-4 156 derivatives on the Hub

finetune: 88quantized: 50adapter: 18

Top variants (40 shown, ranked by downloads)

Variant	Relation	Quant	Params	Band	License	Author	Likes	Downloads
jialinyyzz/humanizer	finetune	GGUF	12.0B	L	apache-2.0	jialinyyzz	2.1k	143k
unsloth/gemma-4-12b	finetune	—	12.0B	L	apache-2.0	unsloth	18	80.2k
Mungert/gemma-4-12B-it-GGUF	quantized	GGUF	—	—	apache-2.0	Mungert	4	30.8k
sneedjak/Adelic-Gemma-4-12B-GGUF	quantized	GGUF	—	—	gemma	sneedjak	6	27.6k
DhruvalLabs/gemma-4-12B-GGUF	quantized	GGUF	—	—	apache-2.0	DhruvalLabs	3	16.8k
mlx-community/gemma-4-12B-4bit	quantized	-4bit	12.0B	L	apache-2.0	mlx-community	12	15k
SurgeFF/AriannaV4	quantized	GGUF	12.0B	L	cc-by-nc-sa-4.0	SurgeFF	1	13.8k
mlx-community/gemma-4-12B-bf16	finetune	-bf16	12.0B	L	apache-2.0	mlx-community	6	8.6k
RedHatAI/gemma-4-12B-it	finetune	—	12.0B	L	apache-2.0	RedHatAI	0	7.9k
mlx-community/gemma-4-12B-8bit	quantized	-8bit	12.0B	L	apache-2.0	mlx-community	6	7.1k
Wuli-art/Gemma-4-for-Qwen-Image-Edit-2511-Prompt-Extend	finetune	—	—	—	apache-2.0	Wuli-art	16	6.2k
mistrjirka/Gemma-4-12B-PXQ-GGUF	quantized	GGUF	—	—	apache-2.0	mistrjirka	0	5.7k
ggml-org/gemma-4-12B-GGUF	quantized	GGUF	—	—	apache-2.0	ggml-org	4	5.5k
professorf/gemma-4-12B-it-gguf	quantized	gguf	—	—	apache-2.0	professorf	1	4.5k
NobodyWho/Google_Gemma4-12B-GGUF	quantized	GGUF	—	—	apache-2.0	NobodyWho	1	3.3k
thebbg/gemma-4-12b-fable-5	quantized	GGUF	—	—	—	thebbg	2	3.2k
mlx-community/gemma-4-12B-nvfp4	quantized	nvfp4	12.0B	L	apache-2.0	mlx-community	4	2.7k
mlx-community/gemma-4-12B-mxfp4	quantized	mxfp4	12.0B	L	apache-2.0	mlx-community	4	2.7k
mlx-community/gemma-4-12B-6bit	quantized	—	12.0B	L	apache-2.0	mlx-community	3	2.3k
mlx-community/gemma-4-12B-mxfp8	quantized	—	12.0B	L	apache-2.0	mlx-community	3	1.5k
LeaderboardModel1/gemma-4-12B-AutoRound-W4A16-RTN	quantized	-W4A16	1.06B	S	—	LeaderboardModel1	2	1.3k
lokeshe09/gemma-4-12b-it-FP8-Dynamic	quantized	-FP8	12.0B	L	apache-2.0	lokeshe09	1	1.3k
majentik/gemma-4-12B-RotorQuant-MLX-8bit	quantized	MLX	12.0B	L	gemma	majentik	1	1.3k
MartinV/clement-mk-gemma-12b	adapter	GGUF	—	—	gemma	MartinV	0	1.1k
lackey43/humanizer	finetune	GGUF	12.0B	L	apache-2.0	lackey43	1	1k
ericthewiz/humanizer	finetune	GGUF	12.0B	L	apache-2.0	ericthewiz	0	1k
Tds810/humanizer	finetune	GGUF	12.0B	L	apache-2.0	Tds810	0	940
MassivDash/Gemma-4-typescript-coder-12b	adapter	GGUF	—	—	mit	MassivDash	0	935
SurgeFF/AriannaV6	quantized	GGUF	12.0B	L	cc-by-nc-sa-4.0	SurgeFF	4	703
lokeshe09/gemma-4-12b-it-INT4-W4A16	quantized	INT4	12.0B	L	apache-2.0	lokeshe09	1	837
Kathir121/humanizer	finetune	GGUF	12.0B	L	apache-2.0	Kathir121	0	853
suryatmodulus/humanizer	finetune	GGUF	12.0B	L	apache-2.0	suryatmodulus	1	791
wezzel98765/gemma-4-12B-oQ4-fp16	quantized	-fp16	12.0B	L	apache-2.0	wezzel98765	1	784
DeepDive404-2/humanizer	finetune	GGUF	12.0B	L	apache-2.0	DeepDive404-2	0	759
LucidCC13105/humanizer	finetune	GGUF	12.0B	L	apache-2.0	LucidCC13105	0	740
ElonCoding/humanizer	finetune	GGUF	12.0B	L	apache-2.0	ElonCoding	0	709
chantra/humanizerth-thai-12B-Q8_0-MLX	quantized	Q8_0	11.9B	L	gemma	chantra	2	595
Joesh1/onca-3.0-12b	finetune	—	12.0B	L	cc-by-nc-4.0	Joesh1	0	694
mlx-community/gemma-4-12B-5bit	quantized	—	12.0B	L	apache-2.0	mlx-community	1	642
npario/humanizer	finetune	GGUF	12.0B	L	apache-2.0	npario	0	680

Repository changelog

6 commits

changed in the last 7 daysOther: 2026-10-24all changes

Repository changelog 6 commits

Date	Commit	SHA
2026-07-15	Update README (#12)	023679e
2026-07-10	Add Technical Report	1dd69cd
2026-06-04	Update max_position_embeddings to full context length (#4)	56820d7
2026-06-03	Update README.md	0c98506
2026-06-03	Update README.md	df8d666
2026-06-03	Preparing for release!	be8e26a

Articles and coverage

42 pieces

Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.

Official4

Gemma 4 12B: A unified, encoder-free multimodal model

blog.google2026-06-03officialauthorityHN front pageHN 1,062 pts · 401 comments584 words

Google releases Gemma 4 12B, an Apache 2.0 licensed model that has crossed 150 million downloads and runs locally on hardware with 16GB of VRAM or unified memory. The encoder-free architecture integrates raw audio signals and vision inputs directly into the LLM backbone, eliminating separate encoders to reduce latency and memory usage. It includes Multi-Token Prediction (MTP) drafters to reduce inference latency while delivering reasoning performance nearing the 26B Mixture of Experts model.written, reviewed, cited

Gemma 4 QAT models: Optimizing compression for mobile and laptop efficiency

blog.google2026-06-05officialauthorityHN front pageabout the gemma 4 familyHN 406 pts · 130 comments714 words

Google released new Gemma 4 checkpoints optimized with Quantization-Aware Training to improve efficiency on edge devices and consumer GPUs. The E2B model achieves a 1GB memory footprint using a mobile-specialized schema that applies targeted 2-bit quantization to token generation layers. Developers can access GGUF formats ready for llama.cpp and MTP QAT checkpoints to preserve inference speedups while running locally.written, reviewed, cited

The mobile-specialized schema pre-calculates activation scaling settings during training to reduce workload on mobile chips.
Targeted 2-bit quantization compresses token generation parts while keeping core reasoning layers at higher precision.
Deploying only needed modalities allows the Gemma 4 E2B text-only model to require less than 1 GB of memory.

Accelerating Gemma 4: faster inference with multi-token prediction drafters

blog.google2026-05-05officialauthorityHN front pageabout the gemma 4 familyHN 687 pts · 330 comments796 words

Google released Multi-Token Prediction (MTP) drafters for the Gemma 4 family under the same open-source Apache 2.0 license as Gemma 4. These drafters deliver up to a 3x speedup for models like the 26B MoE and 31B Dense versions without quality degradation. Benchmarks on hardware such as the NVIDIA RTX PRO 6000 show that increasing batch sizes to 4 to 8 unlocks up to a ~2.2x speedup locally.written, reviewed, cited

The MTP drafter shares the target model's KV cache to avoid recalculating context the larger model has already processed.
For E2B and E4B edge models, an efficient clustering technique in the embedder accelerates generation where logit calculation is a bottleneck.
The 26B mixture-of-experts model presents unique routing challenges at a batch size of 1 on Apple Silicon.

Google releases Gemma 4 open models

deepmind.google2026-04-02officialauthorityHN front pageabout the gemma 4 familyHN 1,812 pts · 474 comments253 words

Analysis6

Show HN: Cactus Hybrid: We taught Gemma 4 to know when it's wrong

github.com2026-07-22authorityHN front pageabout the gemma 4 familyHN 191 pts · 44 comments714 words

Cactus Hybrid introduces on-device models that score every answer with a confidence between 0 and 1 to decide when to hand off to a bigger model. The Gemma 4 E2B Hybrid model matches Gemini 3.1 Flash-Lite performance by routing only 15–55% of queries to the cloud while running the rest locally. It is MIT-licensed and achieves a 0.770 AUROC on MMLU, with a probe trained on zero audio data yet reaching 0.79–0.88 AUROC on four audio benchmarks.written, reviewed, cited

Weights on the meta device crash the confidence read because probes score generations outside the module forward path.
The probe is a patch you compile into the llama.cpp engine rather than a standalone library component.
Transformers version 5.14 and above cause a segfault on this specific checkpoint.

Gemma 4 E2B inference in 700 lines of C

github.com2026-08-27authorityabout the gemma 4 familyHN 18 pts · 0 comments722 words

Fable 5 pushed Gemma 4 to 255 tok/s on WebGPU

xcancel.com2026-06-18about the gemma 4 familyHN 48 pts · 22 comments

Gemma 4 on iPhone

apps.apple.com2026-04-05HN front pageabout the gemma 4 familyHN 868 pts · 234 comments

Google Gemma 4 Runs Natively on iPhone with Full Offline AI Inference

gizmoweek.com2026-04-15HN front pageabout the gemma 4 familyHN 303 pts · 187 comments364 words

Show HN: Determinstic LLM inference for lowest price Gemma 4, with Windows XP

tokendelivery.ai2026-09-12about the gemma 4 familyHN 9 pts · 1 comments

All coverage tracked for this model (42) — append-only, nothing is ever removed

Title	Source	Published	First seen	Signals
Gemma 4 Benchmarks: 31B & 12B Scores, Arena Elo, MMLU Pro \| Gemma4All	gemma4all.com	—	2026-10-07	exa
Show HN: Determinstic LLM inference for lowest price Gemma 4, with Windows XP	tokendelivery.ai	2026-09-12	2026-10-07	hn
Gemma 4 E2B inference in 700 lines of C	github.com	2026-08-27	2026-10-07	hn
Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac	github.com	2026-07-29	2026-10-07	hn
Show HN: Cactus Hybrid: We taught Gemma 4 to know when it's wrong	github.com	2026-07-22	2026-10-07	hn
Running Gemma 4 26B at 5 tokens/sec on a 13-year-old Xeon with no GPU	neomindlabs.com	2026-07-15	2026-10-07	hn
Howardism \| Gemma 4	howardism.dev	2026-07-09	2026-10-07	exa
Gemma 4 Technical Report Deep Dive — Google DeepMind's Open-Weight, Natively Multimodal 2.3B–31B LLMs with an Encoder-Free 12B Unified Design and Built-In Reasoning Mode arXiv:2607.02770, Published 2026-07-02, 300+ Authors, Both Dense and MoE Variants \| Oflight Inc.	oflight.co.jp	2026-07-08	2026-10-07	exa
Show HN: Local privacy-first Microsoft Recall alternative with Gemma 4	github.com	2026-07-04	2026-10-07	hn
Local LLMs on a Ryzen 8700G iGPU: 13-15 tok/s on gemma4, 9-12 on qwen3.6	loomcycle.dev	2026-06-30	2026-10-07	hn
Show HN: Running Gemma-4 26B at 124 tokens/SEC on a CPU, no GPU	apeg.dev	2026-06-30	2026-10-07	hn
Gemma 4 on Cerebras - The Fastest Inference Is Now Multimodal	cerebras.ai	2026-06-30	2026-10-07	hn
I tested Gemma4 12B on my 8GB GPU, now I don't want to go back to smaller models	xda-developers.com	2026-06-19	2026-10-07	hn
Fable 5 pushed Gemma 4 to 255 tok/s on WebGPU	xcancel.com	2026-06-18	2026-10-07	hn
Bonsai: Human->LLM->Web with LLM interface using Gemma4 12B locally on Windows	drive.google.com official	2026-06-09	2026-10-07	hn
Apples to Apples: MLX vs. Llama.cpp for Gemma 4 12B on an M1 16GB	ziraph.com	2026-06-05	2026-10-07	hn
Gemma 4 QAT models: Optimizing compression for mobile and laptop efficiency	blog.google official	2026-06-05	2026-10-07	hn
Gemma 4 12B Developer Guide: Benchmarks & Specs \| Lushbinary	lushbinary.com	2026-06-05	2026-10-07	exa
Google Built Gemma 4 12B Without Multimodal Encoders - Firethering	firethering.com	2026-06-04	2026-10-07	exa
Gemma 4 12B : Run Locally, Fine-Tune, Benchmark Performance	labellerr.com	2026-06-04	2026-10-07	exa
Gemma 4 12B Deep Dive — The Encoder-Free Multimodal LLM That Runs on a 16GB Laptop Under Apache 2.0 (June 3, 2026) \| Oflight Inc.	oflight.co.jp	2026-06-04	2026-10-07	exa
Google's new Gemma 4 12B model is designed to run on any laptop with 16GB of RAM	arstechnica.com	2026-06-03	2026-10-07	hn
A Visual Guide to Gemma 4 12B	newsletter.maartengrootendorst.com	2026-06-03	2026-10-07	hn
Gemma 4 12B: A unified, encoder-free multimodal model	blog.google official	2026-06-03	2026-10-07	hn
Gemma 4 12B: The Developer Guide	developers.googleblog.com official	2026-06-03	2026-10-07	hn, exa
Show HN: I made a Gemma 4 Mac app that names screenshots with local AI	snapname.app	2026-05-31	2026-10-07	hn
Show HN: Local LLM code-generation with Gemma 4 e2B via JSON AST to Clojure	github.com	2026-05-19	2026-10-07	hn
Accelerating Gemma 4: faster inference with multi-token prediction drafters	blog.google official	2026-05-05	2026-10-07	hn
Show HN: Prompt-to-Excalidraw demo with Gemma 4 E2B in the browser (3.1GB)	teamchong.github.io	2026-04-19	2026-10-07	hn
Google Gemma 4 Runs Natively on iPhone with Full Offline AI Inference	gizmoweek.com	2026-04-15	2026-10-07	hn
Show HN: I benchmarked Gemma 4 E2B – the 2B model beat the 12B on multi-turn	aiexplr.com	2026-04-13	2026-10-07	hn
I ran Gemma 4 as a local model in Codex CLI	blog.danielvaughan.com	2026-04-12	2026-10-07	hn
A Visual Guide to Gemma 4	newsletter.maartengrootendorst.com	2026-04-09	2026-10-07	hn
Fitting Gemma 4 (~52 GB) into 12 GB	varjosoft.com	2026-04-08	2026-10-07	hn
Show HN: Gemma 4 Multimodal Fine-Tuner for Apple Silicon	github.com	2026-04-07	2026-10-07	hn
Gemma 4 clocking 12 tokens / SEC on Pixel 7A	twitter.com	2026-04-06	2026-10-07	hn
Gemma 4 on iPhone	apps.apple.com	2026-04-05	2026-10-07	hn
Running Gemma 4 locally with LM Studio's new headless CLI and Claude Code	ai.georgeliu.com	2026-04-05	2026-10-07	hn
April 2026 TLDR Setup for Ollama and Gemma 4 26B on a Mac mini	gist.github.com	2026-04-03	2026-10-07	hn
Google releases Gemma 4 open models	deepmind.google official	2026-04-02	2026-10-07	hn
Gemma 4: Byte for byte, the most capable open models	blog.google official	2026-04-02	2026-10-07	hn
Gemma 4 Architecture and Benchmark Notes \| Sebastian Raschka, PhD	sebastianraschka.com	2026-04-02	2026-10-07	exa

Papers

1 paper

Papers 1

arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).

Gemma 4 Technical Report this model

Gemma Team, Sherif El Abd, Vaibhav Aggarwal +320 · 2026-07-02 · PDF · HF paper page (85 upvotes)

Similar models

3 models

Similar models

Qwen2.5-Omni-7B · 10.7B
same size band (L), same task (any-to-any), permissive license
MiniCPM-o-4_5 · 9.37B
same size band (L), same task (any-to-any), permissive license
MiniCPM-o-2_6 · 8.67B
same task (any-to-any), permissive license

Sources

3 sources

Sources

Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:37 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:06 UTC · Last verified 2026-10-25

Follow gemma-4-12B

Watch this model

/feeds/models/google--gemma-4-12B.xml(RSS 2.0) · every change to gemma-4-12B
/feeds/families/google--gemma-4.xml(RSS 2.0) · every change across the gemma-4 family

Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.