LLaVA 1.6 Vicuna 13B on Replicate API

Name: LLaVA 1.6 Vicuna 13B on Replicate API
Brand: Haotian Liu
Price: 0.10 USD

LLaVA 1.6 · Haotian Liu

Serverless

Last refreshed 2026-04-19. Next refresh: weekly.

Why use LLaVA 1.6 Vicuna 13B on Replicate API?

Replicate API offers LLaVA 1.6 Vicuna 13B with pay-as-you-go pricing at $0.10/1M input tokens. Replicate is a cloud-based platform that enables users to run machine learning models easily and efficiently.

Input / 1M

$0.10

Output / 1M

$0.50

Cache

Not sourced

Batch

Not sourced

Setup recipe

Python + curl

Install

pip install replicate

Auth

export REPLICATE_API_TOKEN=...

Call

import replicate
output = replicate.run(
    "llava-1.6-vicuna-13b",
    input={"prompt": "Hello"}

Model ID

llava-1.6-vicuna-13b

Request example

import replicate

# reads REPLICATE_API_TOKEN from env
# llava-1.6-vicuna-13b format: "owner/model-name" (latest version) or "owner/model-name:version-hash"
output = replicate.run(
    "llava-1.6-vicuna-13b",
    input={"prompt": "Hello"}
)
# Output is a list or generator depending on the model
print("".join(output))

Gotchas

Replicate uses "owner/model-name" format (e.g. "meta/meta-llama-3-8b-instruct") for the latest version, or "owner/model-name:version-sha" to pin to a specific version. The REST endpoint splits owner and model-name into the path: /v1/models/{owner}/{model-name}/predictions.
The examples expect REPLICATE_API_TOKEN; rename it only if your application config maps the new variable.

Pricing

Type	Price (per 1M)
Input tokens	$0.10
Output tokens	$0.50

Capabilities

No model capability flags are currently sourced.

About LLaVA 1.6 Vicuna 13B

LLaVA 1.6 Vicuna 13B is a sophisticated multimodal language model designed to handle multimodal chatbot tasks, integrating text and image processing seamlessly. It features a pre-trained LLM, Vicuna-13B, and a likely CLIP ViT-L/14 vision encoder, linked using a trainable projection matrix, allowing it to comprehend both textual and visual content efficiently. The model offers capabilities such as image captioning, visual question answering, and enhanced reasoning and OCR, with the added advantage of processing high-resolution images up to 672x672 pixels. While it brings notable improvements over its predecessors, it still faces challenges like dependency on diverse training data, computational costs, and inherent biases, necessitating further refinements to maximize its potential across various applications 1238.