v2.2 · 2026-06-12

VISION / MULTIMODAL MODEL — RUNTIME REQUIREMENTS

·CLIP vision encoder with LLaVA multimodal projector; GGUF deployments may require a matching mmproj/projector file or a runtime that bundles it.

·Local vision inference may require a separate projector/mmproj file depending on runtime.

·RAM and VRAM estimates are derived from GGUF/Ollama package size plus conservative overhead; KV cache, image resolution, batch size, GPU offload and runtime support can materially change memory needs.

LLaVA 1.6 Vicuna 13B

llava-1-6-vicuna-13b

VISION-LANGUAGE

LLaVA 1.6 Vicuna 13B is a 13B vision language model profile for local AI planning. This page records provider, license, context, quantization, hardware-fit estimates, setup hints, source links, and caveats so readers do not choose a model by name alone.

PROVIDER

LLaVA / liuhaotian

FAMILY

LLaVA-NeXT / LLaVA 1.6

MODEL TYPE

vision-language

PARAMETERS

13B

MODALITIES

text, image

ARCHITECTURE

LLaVA multimodal adapter over Vicuna 13B v1.5 / Llama architecture, with a CLIP vision encoder and multimodal projector.

CONTEXT WINDOW

4,096 tokens

TRAINING TOKENS

—

RELEASE DATE

2024-01-30

LICENSE & LINKS

Llama 2 Community Licenselicense text

MODEL CARD

huggingface.co/liuhaotian/llava-v1.6-vicuna-13b

HUGGING FACE

liuhaotian/llava-v1.6-vicuna-13b

GGUF REPOSITORIES

cjpais/llava-v1.6-vicuna-13b-gguf community conversion

Community GGUF conversion with Q4_K_M, Q5_K_M and Q8_0 file listings; includes a separate mmproj model file in the repository file listing.

SETUP HINTS

OLLAMA

Run `ollama run llava:13b-v1.6-vicuna-q4_K_M`; Ollama bundles the supported LLaVA runtime path for this package.

LM STUDIO

Use a LLaVA-capable GGUF workflow and ensure the matching multimodal projector is loaded if required.

LLAMA.CPP

Use a current llama.cpp LLaVA/multimodal binary and pass the model plus matching `--mmproj` projector file.

QUANTIZATION FILE SIZES (GGUF)

Q4_K_M

7.87 GB

Q5_K_M

9.23 GB

Q8_0

13.8 GB

Quantization note: Q4_K_M for local image-chat on 16GB+ systems; Q5_K_M if 32GB RAM or ample unified memory is available.

RAM / VRAM ESTIMATES

MIN RAM

16 GB

COMFORTABLE RAM

32 GB

MIN VRAM

10 GB

COMFORTABLE VRAM

16 GB

These are conservative local-inference estimates, not official hardware requirements.

HARDWARE FIT

CPU only (no GPU)

limited

8 GB RAM

16 GB RAM

limited

32 GB RAM

yes

8 GB VRAM

12 GB VRAM

limited

24 GB VRAM

yes

Apple Silicon (unified memory)

yes

Hardware fit values are conservative local-inference estimates based on GGUF size plus runtime and KV-cache overhead. Actual requirements depend on context length, quantization, runtime, GPU offload, and other running apps.

BEST FOR

·Higher-quality local image chat than the 7B LLaVA 1.6 variant

·Commodity workstation multimodal demos

·Research or hobbyist visual question-answering

·Testing larger LLaVA-compatible local runtimes

AVOID IF

·You only have 8GB RAM or 8GB VRAM

·You need production OCR accuracy

·You need permissive Apache/MIT-style commercial licensing

·You need tool calling or long text context

CAVEATS

·Community GGUF conversion is not an official LLaVA artifact.