---
title: Models | Runware Docs
url: https://runware.ai/docs/serverless/models
description: Open models you can run on your own workers, each one deployable without writing any code.
relatedDocuments:
  - https://runware.ai/docs/serverless/examples
  - https://runware.ai/docs/serverless/quickstart
---
## Introduction

Each entry names a model and what serves it, and the build turns that into the source your app runs. **You deploy it as it is**, without writing any code.

You run the app on your own workers, and you pay for the hardware it holds rather than per request.

> [!NOTE]
> These are not the models the inference API serves. To call a model over an API without running anything yourself, use the [inference API](https://runware.ai/docs/models-api/introduction) instead.

All

Text generation

Embeddings

Decisions

Image generation

Video generation

Audio generation

Speech

- [
    
    ### Gemma 4 12B
    
    ![](https://runware.ai/docs/assets/c6494cd0-df37-490c-a3b5-eedadec1d81e_Z21ahyP.webp)Google
    
    - Text generation
    - l40s
    - Apache-2.0
    
    The mid-size Gemma, for when the 31B is more model than the work needs.
    
    ](https://runware.ai/docs/serverless/models/gemma-4-12b)
- [
    
    ### Gemma 4 31B
    
    ![](https://runware.ai/docs/assets/c6494cd0-df37-490c-a3b5-eedadec1d81e_Z21ahyP.webp)Google
    
    - Text generation
    - rtx-pro-6000
    - Apache-2.0
    
    Google's instruction-tuned 31B, served on its own OpenAI-compatible route.
    
    ](https://runware.ai/docs/serverless/models/gemma-4-31b)
- [
    
    ### GLM-4.7 Flash
    
    ![](https://runware.ai/docs/assets/fa229a51-5cf8-4d1b-be22-a89f9645573c_ZnjvBU.webp)Z.ai
    
    - Text generation
    - rtx-pro-6000
    - MIT
    
    A mixture-of-experts model that activates a fraction of its weights per token, so it answers at the speed of a much smaller one.
    
    ](https://runware.ai/docs/serverless/models/glm-4-7-flash)
- [
    
    ### GPT-OSS 120B
    
    ![](https://runware.ai/docs/assets/45bf4bd1-b10a-4da4-9f8b-c08c5d7aed2d_ZAmT2e.webp)OpenAI
    
    - Text generation
    - rtx-pro-6000
    - Apache-2.0
    
    The large open model from OpenAI, 65 GB of MXFP4 weights on a single card.
    
    ](https://runware.ai/docs/serverless/models/gpt-oss-120b)
- [
    
    ### GPT-OSS 20B
    
    ![](https://runware.ai/docs/assets/45bf4bd1-b10a-4da4-9f8b-c08c5d7aed2d_ZAmT2e.webp)OpenAI
    
    - Text generation
    - l40s
    - Apache-2.0
    
    OpenAI's smaller open model, published already quantized, so 21B parameters fit the cheapest card.
    
    ](https://runware.ai/docs/serverless/models/gpt-oss-20b)
- [
    
    ### Nemotron 3 Nano 30B A3B
    
    ![](https://runware.ai/docs/assets/a3744e62-b83d-4e86-af0e-c1976ec7faa1_1WLROk.webp)NVIDIA
    
    - Text generation
    - rtx-pro-6000
    - NVIDIA Open Model License
    
    A mixture-of-experts model with a hybrid attention and state-space architecture, activating a fraction of its weights per token.
    
    ](https://runware.ai/docs/serverless/models/nemotron-3-nano-30b-a3b)
- [
    
    ### Nemotron 3 Nano 4B
    
    ![](https://runware.ai/docs/assets/a3744e62-b83d-4e86-af0e-c1976ec7faa1_1WLROk.webp)NVIDIA
    
    - Text generation
    - l40s
    - NVIDIA Open Model License
    
    NVIDIA's small model, tuned for the hardware it runs on.
    
    ](https://runware.ai/docs/serverless/models/nemotron-3-nano-4b)
- [
    
    ### Nemotron 3.5 Lightning 30B A3B
    
    ![](https://runware.ai/docs/assets/a3744e62-b83d-4e86-af0e-c1976ec7faa1_1WLROk.webp)NVIDIA
    
    - Text generation
    - rtx-pro-6000
    - NVIDIA Open Model License
    
    The latency-tuned member of the Nemotron 3.5 line, a mixture of experts over a hybrid attention and state-space backbone.
    
    ](https://runware.ai/docs/serverless/models/nemotron-3-5-lightning-30b)
- [
    
    ### Qwen3.5 4B
    
    ![](https://runware.ai/docs/assets/cd72ed4c-ef94-4962-a44b-cbe38b3bf6ea_ZUU7vB.webp)Alibaba
    
    - Text generation
    - l40s
    - Apache-2.0
    
    Small enough to keep a worker warm without the bill that usually implies.
    
    ](https://runware.ai/docs/serverless/models/qwen3-5-4b)
- [
    
    ### Qwen3.5 9B
    
    ![](https://runware.ai/docs/assets/cd72ed4c-ef94-4962-a44b-cbe38b3bf6ea_ZUU7vB.webp)Alibaba
    
    - Text generation
    - l40s
    - Apache-2.0
    
    A mid-size model that answers quickly and leaves most of the card for the context.
    
    ](https://runware.ai/docs/serverless/models/qwen3-5-9b)
- [
    
    ### Qwen3.6 35B A3B
    
    ![](https://runware.ai/docs/assets/cd72ed4c-ef94-4962-a44b-cbe38b3bf6ea_ZUU7vB.webp)Alibaba
    
    - Text generation
    - rtx-pro-6000
    - Apache-2.0
    
    A mixture-of-experts model that activates a fraction of its weights per token, in FP8.
    
    ](https://runware.ai/docs/serverless/models/qwen3-6-35b-a3b)
- [
    
    ### Qwen3.8 27B
    
    ![](https://runware.ai/docs/assets/cd72ed4c-ef94-4962-a44b-cbe38b3bf6ea_ZUU7vB.webp)Alibaba
    
    - Text generation
    - l40s
    - Apache-2.0
    
    Alibaba's current flagship open model, in the FP8 weights its authors publish, which halve the memory and let it run on the cheapest card.
    
    ](https://runware.ai/docs/serverless/models/qwen3-8-27b)
- [
    
    ### SmolLM3 3B
    
    ![](https://runware.ai/docs/assets/00458017-c7e5-4426-832b-8d9f8c18a6a9_DC6SI.webp)Hugging Face
    
    - Text generation
    - l40s
    - Apache-2.0
    
    A 3B model trained for reasoning and multilingual use, small enough that a worker comes up in seconds.
    
    ](https://runware.ai/docs/serverless/models/smollm3-3b)
- [
    
    ### BGE Base English
    
    BAAI
    
    - Embeddings
    - l40s
    - MIT
    
    Normalized 768-dimension English embeddings, over a 512-token context.
    
    ](https://runware.ai/docs/serverless/models/bge-base-en)
- [
    
    ### BGE Large English
    
    BAAI
    
    - Embeddings
    - l40s
    - MIT
    
    Normalized 1024-dimension English embeddings, over a 512-token context.
    
    ](https://runware.ai/docs/serverless/models/bge-large-en)
- [
    
    ### BGE Small English
    
    BAAI
    
    - Embeddings
    - l40s
    - MIT
    
    Normalized 384-dimension embeddings from a list of texts in one request, on the smallest card that fits it.
    
    ](https://runware.ai/docs/serverless/models/bge-small-en)
- [
    
    ### BGE-M3
    
    BAAI
    
    - Embeddings
    - l40s
    - MIT
    
    Multilingual embeddings across a hundred languages, for retrieval that is not only English.
    
    ](https://runware.ai/docs/serverless/models/bge-m3)
- [
    
    ### Qwen3 Embedding 0.6B
    
    ![](https://runware.ai/docs/assets/cd72ed4c-ef94-4962-a44b-cbe38b3bf6ea_ZUU7vB.webp)Alibaba
    
    - Embeddings
    - l40s
    - Apache-2.0
    
    Current-generation embeddings from a list of texts in one request, on a model small enough to cost almost nothing to keep warm.
    
    ](https://runware.ai/docs/serverless/models/qwen3-embedding-06b)
- [
    
    ### Qwen3 Embedding 4B
    
    ![](https://runware.ai/docs/assets/cd72ed4c-ef94-4962-a44b-cbe38b3bf6ea_ZUU7vB.webp)Alibaba
    
    - Embeddings
    - l40s
    - Apache-2.0
    
    2560-dimension embeddings over a 40k-token context, the middle size of the family.
    
    ](https://runware.ai/docs/serverless/models/qwen3-embedding-4b)
- [
    
    ### Qwen3 Embedding 8B
    
    ![](https://runware.ai/docs/assets/cd72ed4c-ef94-4962-a44b-cbe38b3bf6ea_ZUU7vB.webp)Alibaba
    
    - Embeddings
    - l40s
    - Apache-2.0
    
    4096-dimension embeddings over a 40k-token context, the largest of the family.
    
    ](https://runware.ai/docs/serverless/models/qwen3-embedding-8b)
- [
    
    ### Clef 27B
    
    ![](https://runware.ai/docs/assets/a11e5453-2045-4154-93bd-6d3160071ed1_24aWPU.webp)Cloudflare
    
    - Decisions
    - rtx-pro-6000
    - Apache-2.0
    
    Typed decisions over text, JSON and images. Send a state and your questions, and every option of every question comes back scored in one pass, with nothing to parse.
    
    ](https://runware.ai/docs/serverless/models/clef-27b)
- [
    
    ### Clef Flash 9B
    
    ![](https://runware.ai/docs/assets/a11e5453-2045-4154-93bd-6d3160071ed1_24aWPU.webp)Cloudflare
    
    - Decisions
    - l40s
    - Apache-2.0
    
    The small, quick Clef. The same typed decisions over text, JSON and images, on a third of the card.
    
    ](https://runware.ai/docs/serverless/models/clef-flash-9b)
- [
    
    ### Clef Omni 30B A3B
    
    ![](https://runware.ai/docs/assets/a11e5453-2045-4154-93bd-6d3160071ed1_24aWPU.webp)Cloudflare
    
    - Decisions
    - rtx-pro-6000
    - Apache-2.0
    
    The Clef that also listens. Send a state with recordings and images alongside it, and your questions come back answered with a probability each.
    
    ](https://runware.ai/docs/serverless/models/clef-omni-30b-a3b)
- [
    
    ### Laya
    
    Convai Innovations
    
    - Decisions
    - l40s
    - Apache-2.0
    
    Typed decisions instead of written answers. Send a state and your questions, and each one comes back answered with a calibrated probability, in English or in a hundred other languages.
    
    ](https://runware.ai/docs/serverless/models/laya-decisions)
- [
    
    ### Fibo 2.0 dev
    
    ![](https://runware.ai/docs/assets/4f68fe9d-851e-4e6a-a6f8-cd47ec4312cb_29oCWp.webp)Bria
    
    - Image generation
    - rtx-pro-6000
    - Bria Fibo license, non-commercial
    
    Bria's 8.1B image model, which reads a structured caption rather than a sentence, so what the image is made of is a field you set rather than a phrase you hope lands. The same weights also edit, with up to five reference images, a mask, and text rendered from fonts it was trained against. Its weights are gated, so this one needs your own Hugging Face token stored as HF\_TOKEN.
    
    ](https://runware.ai/docs/serverless/models/fibo-2-0-dev)
- [
    
    ### Fibo 2.0 Edit Turbo
    
    ![](https://runware.ai/docs/assets/4f68fe9d-851e-4e6a-a6f8-cd47ec4312cb_29oCWp.webp)Bria
    
    - Image generation
    - rtx-pro-6000
    - Bria Fibo license, non-commercial
    
    The distilled Fibo 2.0, which edits in four steps rather than thirty. Same inputs as the full one, up to five reference images and a mask, for a fraction of the time per edit. Its weights are gated, so this one needs your own Hugging Face token stored as HF\_TOKEN.
    
    ](https://runware.ai/docs/serverless/models/fibo-2-0-edit-turbo)
- [
    
    ### FLUX 3 Image Klein
    
    ![](https://runware.ai/docs/assets/07bfd96d-d2df-4d52-aba5-636fb27874be_Z1XEWzy.webp)Black Forest Labs
    
    - Image generation
    - rtx-pro-6000
    - FLUX.3 license
    
    Black Forest Labs' step-distilled FLUX 3, sixteen steps to an image. It expands your prompt into a full scene caption before generating, and returns that caption with the result, so you can see what the image model actually read. A request can name a LoRA adapter to load on top, which is how an adapter trained for FLUX 3 Klein is served. Its weights are gated, so this one needs your own Hugging Face token stored as HF\_TOKEN, and that is also the token a private adapter is fetched with.
    
    ](https://runware.ai/docs/serverless/models/flux-3-klein)
- [
    
    ### FLUX.2 dev
    
    ![](https://runware.ai/docs/assets/07bfd96d-d2df-4d52-aba5-636fb27874be_Z1XEWzy.webp)Black Forest Labs
    
    - Image generation
    - rtx-pro-6000
    - FLUX non-commercial license
    
    The full FLUX.2, not distilled, so it takes a real schedule rather than four steps. Its 113 GB of bfloat16 weights are quantized to int8 as they load, which is what puts it on one card. Its weights are gated, so this one needs your own Hugging Face token stored as HF\_TOKEN.
    
    ](https://runware.ai/docs/serverless/models/flux-2-dev)
- [
    
    ### FLUX.2 Klein 4B
    
    ![](https://runware.ai/docs/assets/07bfd96d-d2df-4d52-aba5-636fb27874be_Z1XEWzy.webp)Black Forest Labs
    
    - Image generation
    - l40s
    - Apache-2.0
    
    The small distilled FLUX, four steps to an image, Apache licensed and light enough to leave a worker warm.
    
    ](https://runware.ai/docs/serverless/models/flux-2-klein-4b)
- [
    
    ### FLUX.2 Klein 9B
    
    ![](https://runware.ai/docs/assets/07bfd96d-d2df-4d52-aba5-636fb27874be_Z1XEWzy.webp)Black Forest Labs
    
    - Image generation
    - l40s
    - FLUX.2 Klein license
    
    The larger distilled FLUX, still four steps. Its weights are gated, so this one needs your own Hugging Face token stored as HF\_TOKEN.
    
    ](https://runware.ai/docs/serverless/models/flux-2-klein-9b)
- [
    
    ### Ideogram 4
    
    ![](https://runware.ai/docs/assets/74d83fcd-ee79-4c06-b923-a7300ca35c17_Z1EMkB1.webp)Ideogram
    
    - Image generation
    - l40s
    - Ideogram license
    
    Ideogram's open release, in the NF4 weights its authors publish for diffusers. Known for typography, so text inside the image comes out readable. Its weights are gated, so this one needs your own Hugging Face token stored as HF\_TOKEN.
    
    ](https://runware.ai/docs/serverless/models/ideogram-4)
- [
    
    ### MiniMax H3 Image
    
    ![](https://runware.ai/docs/assets/b76b19c9-dd5c-4128-9439-7c9ec5088ecc_1aMy3H.webp)MiniMax
    
    - Image generation
    - rtx-pro-6000
    - MiniMax H3 Image community license
    
    MiniMax's image model, six steps and no classifier-free guidance, at four megapixels or at the sixteen MiniMax recommends. Your prompt is read verbatim with no chat template, so a long and complete description is what it works best from, and the first 4096 tokens are what it reads. Strongest at photography, dense scenes and rendered text.
    
    ](https://runware.ai/docs/serverless/models/minimax-h3-image)
- [
    
    ### P-H3-Image
    
    ![](https://runware.ai/docs/assets/9d2bd33f-144d-4a7f-8caf-fed8a74614b7_1Go80H.webp)Pruna AI
    
    - Image generation
    - rtx-pro-6000
    - MiniMax H3 community license
    
    Pruna's image model built on MiniMax H3, eight steps to an 8-megapixel image, no classifier-free guidance. It expands your prompt into the full prompt the model renders best and returns it with the result, so you can read what the image model was actually given. Strongest at photorealism, film and editorial looks, and graphic styles with short quoted text.
    
    ](https://runware.ai/docs/serverless/models/p-h3-image)
- [
    
    ### Qwen Image 2.1
    
    ![](https://runware.ai/docs/assets/cd72ed4c-ef94-4962-a44b-cbe38b3bf6ea_ZUU7vB.webp)Alibaba
    
    - Image generation
    - l40s
    - Qwen license
    
    Alibaba's open image model, strong on rendering text inside the image it generates.
    
    ](https://runware.ai/docs/serverless/models/qwen-image-2-1)
- [
    
    ### Qwen Image 2.1 Turbo
    
    ![](https://runware.ai/docs/assets/cd72ed4c-ef94-4962-a44b-cbe38b3bf6ea_ZUU7vB.webp)Alibaba
    
    - Image generation
    - l40s
    - Qwen license
    
    The distilled Qwen Image, a few steps instead of a full schedule, for when latency matters more than the last detail.
    
    ](https://runware.ai/docs/serverless/models/qwen-image-2-1-turbo)
- [
    
    ### RMBG 2.0
    
    ![](https://runware.ai/docs/assets/4f68fe9d-851e-4e6a-a6f8-cd47ec4312cb_29oCWp.webp)Bria
    
    - Image generation
    - l40s
    - Bria RMBG license, non-commercial
    
    Bria's background remover, which returns the subject on transparency and the mask on its own. It reads the image at a fixed size and gives the mask back at the size you sent, so the result drops straight onto the original. Its weights are gated, so this one needs your own Hugging Face token stored as HF\_TOKEN.
    
    ](https://runware.ai/docs/serverless/models/rmbg-2-0)
- [
    
    ### SAM 2.1 Hiera Large
    
    ![](https://runware.ai/docs/assets/880d16a7-e6d2-4b8b-9fc5-d8bb1c3ef57d_Z27KLNf.webp)Meta
    
    - Image generation
    - l40s
    - Apache 2.0
    
    Meta's Segment Anything 2.1, run without a prompt so it finds every part of an image on its own. It answers with a label map, one PNG the size of the image where each pixel holds the id of the part it belongs to and the smallest part wins wherever parts overlap, so nothing is hidden by what contains it. Each part also comes back with its own area and bounding box, and with its own mask on request. The point grid and the two quality filters are request parameters, so how finely it divides an image is the caller's decision.
    
    ](https://runware.ai/docs/serverless/models/sam-2-1-hiera-large)
- [
    
    ### Z-Image Turbo
    
    ![](https://runware.ai/docs/assets/cd72ed4c-ef94-4962-a44b-cbe38b3bf6ea_ZUU7vB.webp)Alibaba
    
    - Image generation
    - l40s
    - Apache-2.0
    
    A distilled image model that reaches a finished picture in single-digit steps, so a request costs a fraction of a full sampler run.
    
    ](https://runware.ai/docs/serverless/models/z-image-turbo)
- [
    
    ### LTX-2.5
    
    ![](https://runware.ai/docs/assets/5411e7cf-6e95-46e1-bd37-9fd72fc716d1_1X1jk8.webp)Lightricks
    
    - Video generation
    - rtx-pro-6000
    - LTX-2.5 license
    
    Video with its own audio from one pass, and the model picks the length to suit the prompt. Its weights are gated, so this one needs your own Hugging Face token stored as HF\_TOKEN.
    
    ](https://runware.ai/docs/serverless/models/ltx-2-5)
- [
    
    ### MiniMax H3
    
    ![](https://runware.ai/docs/assets/b76b19c9-dd5c-4128-9439-7c9ec5088ecc_1aMy3H.webp)MiniMax
    
    - Video generation
    - rtx-pro-6000
    - MiniMax license
    
    Video with its own stereo audio from one denoising pass, at 768p. This is H3-Base, the part MiniMax published. The context stage their own workflow runs in front of it is not open, so results here are the base model's rather than what their API returns.
    
    ](https://runware.ai/docs/serverless/models/minimax-h3)
- [
    
    ### MiniMax Music 3
    
    ![](https://runware.ai/docs/assets/b76b19c9-dd5c-4128-9439-7c9ec5088ecc_1aMy3H.webp)MiniMax
    
    - Audio generation
    - l40s
    - MiniMax Music 3 community license
    
    A song from a description and its lyrics, sung, in stereo at 44.1 kHz. Leave the lyrics out and the track comes back instrumental.
    
    ](https://runware.ai/docs/serverless/models/minimax-music3)
- [
    
    ### YuE2 3B
    
    M-A-P
    
    - Audio generation
    - l40s
    - CC BY-NC 4.0
    
    A full song from a style and its lyrics, sung, in stereo at 48 kHz. It writes a score first and hands it back as ABC notation, so an edited score can be sent in to render a revision, or a transcribed melody to cover an existing song.
    
    ](https://runware.ai/docs/serverless/models/yue2-3b)
- [
    
    ### Qwen2.5-Omni 7B
    
    ![](https://runware.ai/docs/assets/cd72ed4c-ef94-4962-a44b-cbe38b3bf6ea_ZUU7vB.webp)Alibaba
    
    - Speech
    - rtx-pro-6000
    - Qwen license
    
    A model that listens and answers, in writing or out loud. Send a question, or a recording with a question about it, and ask for the reply to be read aloud.
    
    ](https://runware.ai/docs/serverless/models/qwen-2-5-omni-7b)
- [
    
    ### Qwen3-TTS 0.6B Base
    
    ![](https://runware.ai/docs/assets/cd72ed4c-ef94-4962-a44b-cbe38b3bf6ea_ZUU7vB.webp)Alibaba
    
    - Speech
    - l40s
    - Apache-2.0
    
    Text read aloud in the voice of a recording you supply. A few seconds of clean speech is enough to copy from.
    
    ](https://runware.ai/docs/serverless/models/qwen3-tts-06b-base)
- [
    
    ### Qwen3-TTS 0.6B Custom Voice
    
    ![](https://runware.ai/docs/assets/cd72ed4c-ef94-4962-a44b-cbe38b3bf6ea_ZUU7vB.webp)Alibaba
    
    - Speech
    - l40s
    - Apache-2.0
    
    Text read aloud in one of the voices the model publishes, at a size that starts in seconds.
    
    ](https://runware.ai/docs/serverless/models/qwen3-tts-06b-custom-voice)
- [
    
    ### Qwen3-TTS 1.7B Base
    
    ![](https://runware.ai/docs/assets/cd72ed4c-ef94-4962-a44b-cbe38b3bf6ea_ZUU7vB.webp)Alibaba
    
    - Speech
    - l40s
    - Apache-2.0
    
    Text read aloud in the voice of a recording you supply, at the larger size for closer copies.
    
    ](https://runware.ai/docs/serverless/models/qwen3-tts-17b-base)
- [
    
    ### Qwen3-TTS 1.7B Custom Voice
    
    ![](https://runware.ai/docs/assets/cd72ed4c-ef94-4962-a44b-cbe38b3bf6ea_ZUU7vB.webp)Alibaba
    
    - Speech
    - l40s
    - Apache-2.0
    
    Text read aloud in one of the voices the model publishes, with an optional instruction for how to read it.
    
    ](https://runware.ai/docs/serverless/models/qwen3-tts-17b-custom-voice)
- [
    
    ### Qwen3-TTS 1.7B Voice Design
    
    ![](https://runware.ai/docs/assets/cd72ed4c-ef94-4962-a44b-cbe38b3bf6ea_ZUU7vB.webp)Alibaba
    
    - Speech
    - l40s
    - Apache-2.0
    
    A voice built from a description rather than picked from a list. Say the age, accent, pace and mood, and the model reads your text in it.
    
    ](https://runware.ai/docs/serverless/models/qwen3-tts-17b-voice-design)
- [
    
    ### Voxtral Mini 3B
    
    ![](https://runware.ai/docs/assets/0fdeb675-c722-4f6f-9d82-d30c845e155d_ZIJvS2.webp)Mistral AI
    
    - Speech
    - l40s
    - Apache-2.0
    
    Speech turned into text. Send a recording and get its transcript, with the language detected or named.
    
    ](https://runware.ai/docs/serverless/models/voxtral-mini-3b)

## The deployed app

It behaves like any other Serverless app: it scales to zero on its idle TTL, resumes on the next call, and `apps scale` changes its hardware like anywhere else.

**The weights are pinned to one commit.** An entry records the revision it was tested against, so the same deploy brings the same weights however long it has been listed.

**The server is the one the model's own ecosystem publishes.** A text model runs under vLLM on its own OpenAI-compatible route, from the image vLLM's maintainers release, so there is no Runware layer between your request and the server.