Llama-3.2-11B-Vision-Instruct

meta-llama Llama 3.2 Community License out 2 years ago

10.7B parameters, images and text in, text out.

Verified 2026-10-25 Rank #131 in LLMs · score 94.6 method v1 Sources 3 Downloads 30d 57.5k · #371 of 619 tracked

Can I run it?6.4GB at INT4Estimatefits a 16 GB cardmethod

Can I use it?Llama 3.2 Community LicenseAs 3.1; EU multimodal restriction in the textlicence file2026-10-24

Do people use it?57.5kdownloads, 30 days#371 of 619 we track · below the median+1 likes, downloads −9% this week1,679 likesHugging Face2026-10-24

Worth reading

Picked by rule from 34 tracked pieces: the lab's own post, the most shared, a field report, the latest.

Most discussed · 579 HN pts

Show HN: Llama 3.2 Interpretability with Sparse Autoencoders

The project trained a Sparse Autoencoder on 6.5 Billion activations from Llama 3.2-3B layer 23 over 10 epochs.

github.com 2024-11-21 579 HN pts · 99 comments 5.8k words about the Llama 3.2 family family

Latest · 4 months ago

Show HN: MinLlama – Llama 3.2 inference in ~100 lines of NumPy

A minimal implementation of Llama 3.2 Instruct inference runs in ~100 lines of pure Numpy code.

github.com 2026-06-23 2 HN pts 196 words about the Llama 3.2 family family

25 best of 34 tracked

The 2 picks above, plus these 23. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.

ArticleLlama 3.2: Revolutionizing edge AI and vision with open, customizable modelsai.meta.com · 2024-09-25 · 924 HN pts · official
ArticleLlama 3.2 released: Multimodal, 1B to 90B sizesllama.com · 2024-09-25 · 89 HN pts · official
ArticleI Self-Hosted Llama 3.2 with Coolify on My Home Servergeek.sg · 2024-10-16 · 221 HN pts
ArticleOllama 0.4 is released with support for Meta's Llama 3.2 Vision models locallyollama.com · 2024-11-06 · 182 HN pts
ArticleLlama can now see and run on your device – welcome Llama 3.2huggingface.co · 2024-09-25 · 26 HN pts
ArticleDownload Llama 3.2 1B, 3B, 11B, 90Bllama.com · 2024-09-25 · official
ArticleLlama 3.2 Vision: A Complete Guide, From Core Concepts to a Full Hands-On Lab | atal upadhyayatalupadhyay.wordpress.com · 2026-09-23
ArticleShow HN: Fine-tuned Llama 3.2 3B to match 70B models for local transcriptsbilawal.net · 2025-09-01
ArticleShow HN: Llama 3.2 3B and Keiro Research achieves 85% on SimpleQAkeirolabs.cloud · 2026-03-07
ArticleVerbatim AI – on-device transcription (Whisper) + summaries (Llama 3.2)apps.apple.com · 2026-04-21 · 2 HN pts
CodeLlama 3.2 Releasegithub.com · 2024-09-25 · 12 HN pts
ArticleLoad Llama-3.2 WebGPU in the browser from a local foldersimonwillison.net · 2025-09-09
ArticleRWKV-7 beats Llama 3.2 with 3x fewer training tokens and formally exceeds TC^0ai.gopubby.com · 2026-02-23
CodePydanticAI using Ollama (llama3.2) running locallygithub.com · 2024-12-06
CodeShow HN: Recall – open-source local file organizer using Llama 3.2 and Ollamagithub.com · 2025-12-10
CodeImplementing the Llama 3.2 1B and 3B Architectures from Scratchgithub.com · 2024-10-05
CodeShow HN: MLX.zig – Llama 3.2 in Ziggithub.com · 2025-03-14
ArticleOpen Source wireframe to app powered by Llama 3.2 Visionnapkins.dev · 2024-09-25
ArticleShow HN: Finetune Llama 3.2 Vision in a Colabcolab.research.google.com · 2024-11-21
DiscussionLlama 3.2: Lightweight models for edge devicestwitter.com · 2024-09-25
ArticleLlama 3.2 Vision 11B Review: 50-Image Multimodal Test | BestLLMforbestllmfor.com · 2026-05-24
DiscussionLlama 3.2 1B and 3B on-device modelstwitter.com · 2024-09-25
ArticleMeta Llama 3.2 Review — First Multimodal Llama, Built for the Edge — ChatForestchatforest.com · 2026-05-13

Show all 23

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: 20 quantized builds and 100 Spaces.

As a service0no API provider listed
On own hardware20quantized builds
Built on473161 fine-tunes · 308 adapters · 4 merges
In apps100+Hugging Face Spaces

Top Spaces running it: huggingface-projects/llama-3.2-vision-11B ♥ 389 · algorithmicsuperintelligence/OptiLLM ♥ 27 · build-small-hackathon/pawmap ♥ 11 · NiansuhAI/Main ♥ 10

Will it fit?

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

Fits a 16 GB card at INT4

About 6.4 GB at INT4, 13 GB at FP8, 26 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.

10 dated articles and 0 repo commits so far

Teal is the weights (2024-09-18), amber triangles are 10 dated articles (each links to it).

What people built on it

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

493 derivatives on the Hub, most of them adapters

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

GGUF leafspark/Llama-3.2-11B-Vision-Instruct-GGUF for llama.cpp, Ollama, LM Studio 322k downloads MLX mlx-community/Llama-3.2-11B-Vision-Instruct-8bit for Apple silicon 28.6k downloads FP8 amd/Llama-3.2-11B-Vision-Instruct-FP8-KV for vLLM, SGLang 6.3k downloads

All variants →

Reference

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

Use-case evidence

2 use cases

What it is good at 2 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

17 fields

Listed in LLMs

Facts

Llama-3.2-11B-Vision-Instruct has 10.7B parameters. (safetensors metadata, observed 2026-10-24)
License: Llama 3.2 Community License (friction: attribution). As 3.1; EU multimodal restriction in the text. (license file, observed 2026-10-24)
Created 2024-09-18, last modified 2024-12-04. (HF API, observed 2026-10-24)
Access is gated on Hugging Face (manual). (HF API, observed 2026-10-24)

Pipeline tag	image-text-to-text
Library	transformers
Architectures	MllamaForConditionalGeneration
Model type	mllama
Params by dtype	BF16: 10,670,220,835
Size band	L
MoE	no
Gated	manual
Papers	arXiv:2204.05149
License	Llama 3.2 Community License (llama3.2)
Hub snapshot	2026-10-25

Servability

vLLM · SGLang

Servability

vLLM: not in registry · MllamaForConditionalGeneration · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · MllamaForConditionalGeneration · 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 Llama-3.2-11B-Vision-Instruct: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)	26 GB · Single high-end GPU (<80 GB)
FP8 (1 byte/param)	13 GB · Single GPU (<24 GB)
INT4 (0.5 bytes/param)	6.4 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: meta-llama/Llama-3.2 493 derivatives on the Hub

quantized: 20finetune: 161adapter: 308merge: 4

Top variants (40 shown, ranked by downloads)

Variant	Relation	Quant	Params	Band	License	Author	Likes	Downloads
unsloth/Llama-3.2-11B-Vision-Instruct-bnb-4bit	quantized	-bnb	11.0B	L	llama3.2	unsloth	86	1.2M
unsloth/Llama-3.2-11B-Vision-Instruct-unsloth-bnb-4bit	quantized	-bnb	10.7B	L	llama3.2	unsloth	30	904.4k
unsloth/Llama-3.2-11B-Vision-Instruct	finetune	—	10.7B	L	llama3.2	unsloth	89	683.1k
leafspark/Llama-3.2-11B-Vision-Instruct-GGUF	quantized	GGUF	—	—	llama3.2	leafspark	74	322k
Xkev/Llama-3.2V-11B-cot	finetune	—	10.7B	L	apache-2.0	Xkev	158	76.4k
SeanScripts/Llama-3.2-11B-Vision-Instruct-nf4	quantized	-nf4	10.8B	L	llama3.2	SeanScripts	13	39.3k
omkarthawakar/LlamaV-o1	finetune	—	10.7B	L	apache-2.0	omkarthawakar	95	27.6k
mlx-community/Llama-3.2-11B-Vision-Instruct-8bit	finetune	-8bit	10.7B	L	llama3.2	mlx-community	9	28.6k
mlx-community/Llama-3.2-11B-Vision-Instruct-abliterated	finetune	—	10.7B	L	llama3.2	mlx-community	7	28.3k
mlx-community/Llama-3.2-11B-Vision-Instruct-4bit	finetune	-4bit	10.7B	L	llama3.2	mlx-community	8	20.7k
royleibov/Llama-3.2-11B-Vision-Instruct-ZipNN-Compressed	finetune	—	—	—	llama3.2	royleibov	4	17.2k
pbatra/Llama-3.2-11B-Vision-Instruct-GGUF	quantized	GGUF	—	—	llama3.2	pbatra	5	14.5k
EnlistedGhost/Llama-3.2-11B-Vision-Instruct-OLLAMA	quantized	GGUF	—	—	llama3.2	EnlistedGhost	2	11.6k
AdaptLLM/biomed-Llama-3.2-11B-Vision-Instruct	finetune	—	10.7B	L	llama3.2	AdaptLLM	2	10.1k
Sci-fi-vy/Llama-3.2-11B-Vision-Instruct-GGUF	quantized	GGUF	—	—	llama3.2	Sci-fi-vy	1	7.9k
amd/Llama-3.2-11B-Vision-Instruct-FP8-KV	quantized	-FP8	10.7B	L	llama3.2	amd	1	6.3k
lamm-mit/Cephalo-Llama-3.2-11B-Vision-Instruct-128k	finetune	—	10.7B	L	apache-2.0	lamm-mit	6	4.7k
mlx-community/Llama-3.2-11B-Vision-Instruct-abliterated-4-bit	finetune	—	10.7B	L	llama3.2	mlx-community	1	4.7k
fahadh4ilyas/llama3.2-11B-Vision-Instruct-INT4-GPTQ	quantized	INT4	10.7B	L	llama3.2	fahadh4ilyas	1	4.6k
OPEA/Llama-3.2-11B-Vision-Instruct-int4-sym-inc	quantized	int4	3.11B	M	llama3.2	OPEA	2	4k
cragtmp/taskgt705-450	adapter	—	—	—	—	cragtmp	0	3.9k
mlx-community/Llama-3.2-11B-Vision-Instruct-abliterated-8-bit	finetune	—	10.7B	L	llama3.2	mlx-community	1	2.9k
Kendamarron/Llama-3.2-11B-Vision-Instruct-Swallow-8B-Merge	merge	—	10.7B	L	llama3.2	Kendamarron	3	2.1k
fbaldassarri/meta-llama_Llama-3.2-11B-Vision-Instruct-OpenVino	finetune	—	—	—	llama3.2	fbaldassarri	1	2.2k
Kun-Xiang/AtomThink-Llama-3.2-Vision-11B	finetune	—	10.7B	L	—	Kun-Xiang	2	2k
monirehebrahimi/fine-tuned-visionllama	adapter	—	—	—	—	monirehebrahimi	0	1.7k
EricB/Llama-3.2-11B-Vision-Instruct-UQFF	quantized	—	—	—	—	EricB	16	880
zhangsongbo365/Llama-3.2V-11B-cot-nf4	quantized	-nf4	11.0B	L	apache-2.0	zhangsongbo365	0	1.7k
AdaptLLM/remote-sensing-Llama-3.2-11B-Vision-Instruct	finetune	—	10.7B	L	llama3.2	AdaptLLM	4	1.3k
OrchardPair/Llama-3.2-11B-Vision-Instruct-4bit	finetune	-4bit	10.7B	L	llama3.2	OrchardPair	0	1.4k
fairy322/Llama-3.2-11B-Vision-Instruct-abliterated-8-bit	finetune	—	10.7B	L	llama3.2	fairy322	0	1.4k
Kadins/Llama-3.2-Vision-chinese-lora	adapter	—	—	—	apache-2.0	Kadins	12	682
cragtmp/t3r	adapter	—	—	—	—	cragtmp	0	1.1k
Cannae-AI/MedicalLlama3.2-vision-11B-IT	finetune	—	10.7B	L	apache-2.0	Cannae-AI	12	457
Aekanun/thai-handwriting-llm	finetune	—	—	—	apache-2.0	Aekanun	7	690
AdaptLLM/food-Llama-3.2-11B-Vision-Instruct	finetune	—	10.7B	L	llama3.2	AdaptLLM	6	676
OPEA/Llama-3.2-11B-Vision-Instruct-qvision-int4-sym-inc	quantized	int4	2.43B	M	—	OPEA	1	893
maum-ai/Llama-3.2-MAAL-11B-Vision-v0.1	finetune	—	10.7B	L	llama3.2	maum-ai	1	881
vector-institute/Llama3.2-Multimodal-Newsmedia-Bias-Detector	finetune	—	11.0B	L	cc	vector-institute	4	505
HuanjinYao/Mulberry_llama_11b	finetune	—	10.7B	L	apache-2.0	HuanjinYao	1	636

Articles and coverage

34 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.

Official2

Llama 3.2: Revolutionizing edge AI and vision with open, customizable models

ai.meta.com2024-09-25officialauthorityHN front pageabout the Llama 3.2 familyHN 924 pts · 328 comments2,460 words

Meta releases Llama 3.2, featuring small and medium-sized vision LLMs (11B and 90B) alongside lightweight text-only models (1B and 3B) designed for edge and mobile devices. The 11B and 90B models serve as drop-in replacements for their corresponding text model equivalents while the 1B and 3B models support a context length of 128K tokens and are enabled on day one for Qualcomm and MediaTek hardware. Both pre-trained and aligned vision models are available for fine-tuning using torchtune and local deployment via torchchat, with weights based on BFloat16 numerics.written, reviewed, cited

The 11B and 90B vision models require an entirely new model architecture that supports image reasoning.
Developers can keep all text-only capabilities intact because the adapter training intentionally did not update the language-model parameters.
The 1B and 3B models are created by using structured pruning in a single shot manner from the Llama 3.1 8B.

Llama 3.2 released: Multimodal, 1B to 90B sizes

llama.com2024-09-25officialauthorityabout the Llama 3.2 familyHN 89 pts · 19 comments42 words

Analysis8

Show HN: Llama 3.2 Interpretability with Sparse Autoencoders

github.com2024-11-21authorityHN front pageabout the Llama 3.2 familyHN 579 pts · 99 comments5,768 words

The project trained a Sparse Autoencoder on 6.5 Billion activations from Llama 3.2-3B layer 23 over 10 epochs. Data capture processed 25 million sentences totaling 4TB of raw activation data compressed to 3.2TB.generated, cited

I Self-Hosted Llama 3.2 with Coolify on My Home Server

geek.sg2024-10-16HN front pageabout the Llama 3.2 familyHN 221 pts · 90 comments2,289 words

A guide details self-hosting Llama 3.2 on a home server using Coolify and Ollama. The author achieved a 10x reduction in inference time by enabling GPU acceleration on an NVIDIA GeForce RTX 2060 using the CUDA toolkit and nvidia-container-toolkit. The setup involves installing Ubuntu 24, configuring Cloudflare Tunnels, and managing CUDA toolkits.written, reviewed, cited

Anyone with physical access can read your data if you avoid LVM or disk encryption for smoother reboots.
Cloudflare provides free SSL certificates only for wildcard domains at the top level, not on subdomains.
The Ollama service initially runs on the CPU until the NVIDIA Container Toolkit is installed and configured.

Show HN: MinLlama – Llama 3.2 inference in ~100 lines of NumPy

github.com2026-06-23authorityabout the Llama 3.2 familyHN 2 pts · 0 comments196 words

A minimal implementation of Llama 3.2 Instruct inference runs in ~100 lines of pure Numpy code. The project supports Python ≥ 3.10 and includes backends for PyTorch and Jax alongside the primary Numpy version. Ingredients in Llama 3.2 include RoPE with the Llama-3 frequency scaling, GQA, RMSNorm, SwiGLU MLP, and a shared unembedding / embedding matrix.written, reviewed, cited

Ollama 0.4 is released with support for Meta's Llama 3.2 Vision models locally

ollama.com2024-11-06HN front pageabout the Llama 3.2 familyHN 182 pts · 25 comments202 words

Llama can now see and run on your device – welcome Llama 3.2

huggingface.co2024-09-25authorityabout the Llama 3.2 familyHN 26 pts · 1 comments3,434 words

Llama 3.2 Vision: A Complete Guide, From Core Concepts to a Full Hands-On Lab | atal upadhyay

atalupadhyay.wordpress.com2026-09-23about the Llama 3.2 family3,947 words

Verbatim AI – on-device transcription (Whisper) + summaries (Llama 3.2)

apps.apple.com2026-04-21about the Llama 3.2 familyHN 2 pts · 1 comments

Llama 3.2 Release

github.com2024-09-25authorityabout the Llama 3.2 familyHN 12 pts · 0 comments

All coverage tracked for this model (34) — append-only, nothing is ever removed

Title	Source	Published	First seen	Signals
meta / llama-3.2-11b-vision-instruct	docs.api.nvidia.com	—	2026-10-07	exa
Llama 3.2 11B: Benchmarks, Pricing and Specs \| llmboard.ai	llmboard.ai	—	2026-10-07	exa
Llama 3.2 11B Vision Instruct - Benchmarks, Pricing & API Access \| AI Stats	ai-stats.phaseo.app	—	2026-10-07	exa
Llama 3.2 Guide: How It Works, Use Cases & More	datacamp.com	—	2026-10-07	exa
Llama 3.2 Vision: A Complete Guide, From Core Concepts to a Full Hands-On Lab \| atal upadhyay	atalupadhyay.wordpress.com	2026-09-23	2026-10-07	exa
Show HN: MinLlama – Llama 3.2 inference in ~100 lines of NumPy	github.com	2026-06-23	2026-10-07	hn
Llama 3.2 Vision 11B Review: 50-Image Multimodal Test \| BestLLMfor	bestllmfor.com	2026-05-24	2026-10-07	exa
Meta Llama 3.2 Review — First Multimodal Llama, Built for the Edge — ChatForest	chatforest.com	2026-05-13	2026-10-07	exa
Verbatim AI – on-device transcription (Whisper) + summaries (Llama 3.2)	apps.apple.com	2026-04-21	2026-10-07	hn
Show HN: Llama 3.2 3B and Keiro Research achieves 85% on SimpleQA	keirolabs.cloud	2026-03-07	2026-10-07	hn
RWKV-7 beats Llama 3.2 with 3x fewer training tokens and formally exceeds TC^0	ai.gopubby.com	2026-02-23	2026-10-07	hn
Show HN: Recall – open-source local file organizer using Llama 3.2 and Ollama	github.com	2025-12-10	2026-10-07	hn
Load Llama-3.2 WebGPU in the browser from a local folder	simonwillison.net	2025-09-09	2026-10-07	hn
Show HN: Fine-tuned Llama 3.2 3B to match 70B models for local transcripts	bilawal.net	2025-09-01	2026-10-07	hn
LLaMA 3.2-Vision-Instruct: A Layer-Wise Guide to Attention, Embeddings, and Multimodal Reasoning	guttikondaparthasai.medium.com	2025-04-03	2026-10-07	exa
Show HN: MLX.zig – Llama 3.2 in Zig	github.com	2025-03-14	2026-10-07	hn
Run Llama 3.2 3B on your phone	apps.apple.com	2025-01-18	2026-10-07	hn
I built this robot and gave it a llama3.2 brain	twitter.com	2025-01-14	2026-10-07	hn
PydanticAI using Ollama (llama3.2) running locally	github.com	2024-12-06	2026-10-07	hn
Show HN: Llama 3.2 Interpretability with Sparse Autoencoders	github.com	2024-11-21	2026-10-07	hn
Show HN: Finetune Llama 3.2 Vision in a Colab	colab.research.google.com	2024-11-21	2026-10-07	hn
LLama3.2-vision as almost perfect OCR?	demo.doctractor.com	2024-11-18	2026-10-07	hn
Ollama 0.4 is released with support for Meta's Llama 3.2 Vision models locally	ollama.com	2024-11-06	2026-10-07	hn
I Self-Hosted Llama 3.2 with Coolify on My Home Server	geek.sg	2024-10-16	2026-10-07	hn
Implementing the Llama 3.2 1B and 3B Architectures from Scratch	github.com	2024-10-05	2026-10-07	hn
Llama 3.2 Complete Guide: Lightweight & Vision Models	llamaimodel.com	2024-09-26	2026-10-07	exa
Llama can now see and run on your device – welcome Llama 3.2	huggingface.co	2024-09-25	2026-10-07	hn
Open Source wireframe to app powered by Llama 3.2 Vision	napkins.dev	2024-09-25	2026-10-07	hn
Llama 3.2 1B and 3B on-device models	twitter.com	2024-09-25	2026-10-07	hn
Llama 3.2: Lightweight models for edge devices	twitter.com	2024-09-25	2026-10-07	hn
Llama 3.2 Release	github.com	2024-09-25	2026-10-07	hn
Llama 3.2: Revolutionizing edge AI and vision with open, customizable models	ai.meta.com official	2024-09-25	2026-10-07	hn
Download Llama 3.2 1B, 3B, 11B, 90B	llama.com official	2024-09-25	2026-10-07	hn
Llama 3.2 released: Multimodal, 1B to 90B sizes	llama.com official	2024-09-25	2026-10-07	hn

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).

The Carbon Footprint of Machine Learning Training Will Plateau, Then Shrink cited in card

David Patterson, Joseph Gonzalez, Urs Hölzle +7 · 2022-04-11 · PDF · HF paper page (13 upvotes)

Similar models

3 models

Similar models

Qwen3.5-9B · 9.65B
same size band (L), same task (image-text-to-text), more providers, permissive license
GLM-4.6V-Flash · 10.3B
same size band (L), same task (image-text-to-text), more providers, permissive license
Qwen3-VL-8B-Instruct · 8.77B
same task (image-text-to-text), more providers, permissive license

Sources

3 sources

Sources

Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:34 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:01 UTC · Last verified 2026-10-25

Follow Llama-3.2-11B-Vision-Instruct

Watch this model

/feeds/models/meta-llama--Llama-3.2-11B-Vision-Instruct.xml(RSS 2.0) · every change to Llama-3.2-11B-Vision-Instruct
/feeds/families/meta-llama--Llama-3.2.xml(RSS 2.0) · every change across the Llama-3.2 family

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