# Multimedia Solutions | Unlock Insights Across Text & Images with Fireworks

> Turn unstructured data into actionable insights with Fireworks. Fine-tuned models extract, classify, and summarize documents, images, and audio in real time, enabling faster, smarter decisions at enterprise scale

## Multimodal AI for images and documents

Fireworks is the open-model platform for multimodal. Send images through an OpenAI-compatible API, with prompt caching that can cut time to first token (TTFT) by up to 80%. Then fine-tune on your own document types.

[Talk to our team](https://fireworks.ai/contact) · [Get started](https://app.fireworks.ai/signup)

## How multimodal runs on Fireworks

You'll use inference and training for most image work.

[Talk to our team](https://fireworks.ai/contact) · [Read the docs](https://docs.fireworks.ai)

- **Inference**: Call vision models on serverless to start, or move to an on-demand deployment for fine-tuned LoRAs; serverless bills per token, and cached input costs less. · [Vision docs](https://docs.fireworks.ai)
- **Training**: When a base model misreads your forms, run vision SFT in the same managed job flow as text SFT and serve the LoRA on an on-demand deployment. · [Fine-tuning docs](https://docs.fireworks.ai)

## Why use Fireworks for multimodal?

Own the vision model you tune on your documents

- **Lower cost**: Healthcare and insurance firms ran 100x cheaper, 1.5x faster than GPT-4o (2024).
- **Own your model**: AlliumAI runs dozens of fine-tuned LoRA models and pays only for usage (2024).
- **Launch partner**: Fireworks has been Moonshot's launch partner for Kimi K2.5, K2.6 and K2.7.
- **OpenAI SDK**: Keep the OpenAI SDK and the vision request format you already use.

## How to set up multimodal on Fireworks

How it works

[Read the docs](https://docs.fireworks.ai)

- **Step 1: Pick a vision model**: Start with Kimi K2.6, the top pick for vision and document understanding in the Fireworks recommended-models guide.
- **Step 2: Swap the base URL and key**: The request format matches OpenAI's vision API, so the OpenAI SDK works once you change the base URL and API key.
- **Step 3: Send images by URL or base64**: URLs must be under 5MB and download within 1.5 seconds; base64 must total under 10MB. Use PNG, JPEG, GIF, BMP, TIFF or PPM.
- **Step 4: Convert PDFs to page images**: Vision models don't accept PDFs natively. Render one image per page and send up to 30 images per request.
- **Step 5: Cut latency per request**: Use prompt caching, downsize images, use URLs in long chats, put static instructions first and add image metadata.

## Which models can I use for multimodal?

Vision models

These image and document models come from the Fireworks recommended-models guide, last updated August 2026. Each one takes image input.

[See all models](https://fireworks.ai/models)

- **[Kimi K2.6](https://fireworks.ai/models/fireworks/kimi-k2p6)**: Moonshot AI · 262,144 Context · Vision
- **[Kimi K2.5](https://fireworks.ai/models/fireworks/kimi-k2p5)**: Moonshot AI · 262,144 Context · Vision
- **[Qwen3.7 Plus](https://fireworks.ai/models/fireworks/qwen3p7-plus)**: Qwen · Vision
- **[Step-3.7-Flash-NVFP4](https://fireworks.ai/models/fireworks/step-3p7-flash-nvfp4)**: StepFun · 262,144 Context · Vision

## How to prepare documents for multimodal

Guide

### Turning PDFs into images

Vision models on Fireworks don't read PDF files. You render each page as an image and send the pages as image parts, within the per-request cap in step 4. A 12-page contract becomes 12 images. If you need answers across thousands of stored PDFs, that retrieval job belongs on Enterprise RAG.

### Estimating image token cost

Every page image bills as input tokens. Most common resolutions run 1,000–2,500 tokens per image; in the docs example, a 1024×1024 image on Qwen2.5 VL is 1,369 tokens. Render pages at the smallest size your model still reads. Multiply tokens per page by the input price in the Models section.

### Fine-tuning on your document types

When a base model mislabels your forms, train it on your own pages. Training images must be base64 data URIs. You get a LoRA that runs only on an on-demand deployment. Live merge matches base-model latency and throughput, while multi-LoRA often raises TTFT by 10–30%.

## Learn more about multimodal

Resources

[Browse the docs](https://docs.fireworks.ai)

- **[MiniMax M3 is live: long context + native multimodality at 1/20th the price](https://fireworks.ai/blog/minimax-m3)**: 12 JUN, 2026 · Model release
- **[Accelerate your vision pipelines with the new NVIDIA Nemotron Nano 2 VL model on Fireworks](https://fireworks.ai/blog/nvidia-nemotron-nano-2-vl)**: 27 OCT, 2025 · Model release
- **[Introducing vision-language model fine-tuning: tailor VLMs to your domain](https://fireworks.ai/blog/vlm-fine-tuning)**: 29 JUL, 2025 · Announcement

## Frequently asked questions about multimodal

FAQ

[Talk to our team](https://fireworks.ai/contact)

### What's involved in adding vision to my app?

Fireworks takes OpenAI-format vision requests, so your OpenAI SDK code carries over. Add image_url parts to your messages and convert PDFs to page images first.

### What's the fastest API for vision models?

Fireworks hasn't published a vision latency benchmark, so we won't rank ourselves. Test your own images at production load and log TTFT, throughput, model, hardware and date.

### Can I host multimodal models serverless?

Yes. Vision models run on serverless, and the Model Library's Vision filter lists them. Fine-tuned LoRA models need an on-demand deployment.

### Can I send PDFs to a vision model?

Not as files. Convert each page to an image and send the pages as image parts. The guide above covers conversion and token cost.

### What does an image cost on Fireworks?

Images bill as input tokens at your model's serverless rate. [Batch inference costs 50% of serverless pricing](https://docs.fireworks.ai/serverless/pricing).

### How does Fireworks handle my image data?

You choose how each image travels: pass an image URL, or embed it as base64 inline in the request. Open models run with Zero Data Retention by default.

## Send your first image request

Send images through an OpenAI-compatible API, then fine-tune on your own document types.

[Get started](https://app.fireworks.ai/signup/) · [Talk to an expert](https://fireworks.ai/contact)
