llmrun

Best local LLMs for vision

Run local LLMs that handle images, screenshots, and docs across your GPU or Mac. We rank them using the llmrun Composite Score and highlight our top choice for every VRAM tier.

Best for vision on

8 GB12 GB16 GB24 GB32 GB48 GB+

DeepSeek V4.1 Flash 1017

DeepSeek · Q4_K_M · needs 458 GB · up to 1M context · runs on about 7× H100 SXM 80 GB · from $18.83/hr in the cloud

No memory limit: the best model, whatever memory it needs.

How to run →

All 147 vision models, no memory limit ↓

Also worth a look

Kimi K3

Q4_K_M · needs 1.7 TB · about 3 nodes of 8× H100 SXM 80 GB · 1M ctx

Best vision model by memory

Highest llmrun Composite Score that fits each size with at least 16K context and 10 t/s, at a Q3 or better quant. Speed is decode tokens per second on that size's reference GPU. The last row has no memory limit: the best model whatever it needs, on the GPU plan shown. Full ranking

Vision models, no memory limit147

Ranked by Composite Score, models without one after. Memory is what the listed quant needs; speed is shown only for models that fit one 48 GB GPU.

147 models

Featured only

Score: llmrun Score, 100 = average model. Ordered by Score. How ranking works

LLM models ranked by Score, recency and Fit
ModelVRAMScore
ZDTaichu5.0 9B
Q4_K_M·96.6 t/s
9.2 GB
Molmo2 4B
Q4_K_M·183.5 t/s
4.7 GB
Qwen3.5 0.8B
Q4_K_M·717.2 t/s
0.9 GB
Youtu Parsing Omni
BF16·55.8 t/s
12.7 GB
EXAONE 4.5 33B
Q4_K_M·29.1 t/s
25.2 GB
Kimi VL A3B Instruct
Q4_K_M·201.4 t/s
10.7 GB
Agnes 3.0 Flash
Q4_K_M·30.2 t/s
24.2 GB
Chandra Ocr 2
Q4_K_M·171.0 t/s
3.8 GB
InternVL3 5 1B
Q4_K_M·891.4 t/s
2.2 GB
SmolVLM2 500M Video Instruct
< 16K context fits
Q4_K_M·904.3 t/s
0.9 GB
InternVL3 5 8B
Q4_K_M·110.8 t/s
8.2 GB
Qwen3 VL 8B Instruct
Q4_K_M·106.5 t/s
8.0 GB
MiniCPM V 4
Q4_K_M·224.5 t/s
3.1 GB
DeepSeek OCR
< 16K context fits
Q4_K_M·502.0 t/s
2.8 GB
Llama 3.1 Nemotron Nano VL 8B V1
Q4_K_M·108.3 t/s
8.4 GB

How much memory do you need for a vision model?

On 8 GB the best vision model is Qwen3.5 9B (score 99). At 12 GB it is Gemma 4 26B A4B IT (score 106). From 16 to 32 GB it is Qwen3.8 27B (score 122). With no memory limit, the best is DeepSeek V4.1 Flash 1017 (score 133). Add a bit more breathing room to your setup when using a vision model with an image encoder alongside the language model; each image loads as extra tokens in the context. Allocate slightly more space than you would for a text-only model of equal size.

Frequently Asked Questions

What is the best local LLM for vision?

On 24 GB, Qwen3.8 27B is the top pick with a score of 122, running at about 38 t/s on RTX 4090. It fits up to 64K tokens of context there. On 8 GB, the best fit is Qwen3.5 9B. With no memory limit, the best is DeepSeek V4.1 Flash 1017. Open a model to see which GPUs and Macs run it.

Can I run a vision model on a laptop or an 8 GB GPU?

Yes. On 8 GB the best pick is Qwen3.5 9B (Q4_K_M, 6.6 GB), which runs at about 46 t/s on RTX 5060. Laptops with 8 to 16 GB of memory can run smaller models in the same way.

Can local models read screenshots and documents?

Vision-language models take an image along with your prompt to answer questions, pull out text, or describe charts. Whether support works depends on the runtime; however, you can run them in Ollama or LM Studio.

Do images need extra memory?

The image encoder bumps up the model size. Each picture turns into hundreds of tokens within the context window, swelling the cache. Bigger images demand even more.