
When you're getting started with an AI project, choosing the right model is only half the battle. You also need to configure environments and install dependencies while connecting the tools that make your workflows run smoothly. Setting things up can quickly become a bottleneck.

Prebuilt templates simplify the process – and **Vast.ai** gives you plenty to choose from!

Our [Model Library](https://vast.ai/model-library) offers a model-first way to deploy AI, with templates that launch ready-to-run environments for some of the most-used AI tools today. Vast.ai also offers [lower-level templates](https://cloud.vast.ai/templates) for frameworks and inference engines like PyTorch and vLLM for more customizable setups.

Each pre-configured template is organized by modality and use case so you can quickly find what you need – and you retain full custom control over the underlying runtime, parameters, and infrastructure.

### Popular Templates in the Vast.ai Model Library

From image, text, and audio-video generation to multimodal AI workflows, the following are a few of the most popular templates being deployed on Vast.ai right now.

#### 1. Flux.2-Dev – Rectified Flow Image Generation

[**FLUX.2-Dev**](https://vast.ai/model/flux.2-dev) is a 32-billion parameter rectified flow transformer built for advanced text-to-image generation and editing. The model processes text prompts alongside optional reference images, so you can modify existing images or combine multiple visual references without additional fine-tuning.

Its ability to maintain consistent characters, objects, and styles across generations makes it ideal for character design, visual storytelling with scene continuity, digital art with precise style control, and design prototyping.

#### 2. LTX-2 – Native Audio-Video Generation in a Single Workflow

[**LTX-2**](https://vast.ai/model/ltx-2) is a 19-billion parameter Diffusion Transformer (DiT) model that generates audio and video together in a single pass. It creates synchronized audiovisual output from text prompts, images, and other inputs – keeping dialogue, motion, and ambient sound aligned throughout the clip.

One of LTX-2's biggest advantages is its light and iterative workflow. Since the model generates video in a compressed latent space before converting it to full resolution, it supports faster iteration and more efficient memory usage than many traditional video models.

#### 3. WAN 2.2 – Efficient Text-to-Video and Image-to-Video Generation

[**WAN 2.2**](https://vast.ai/article/wan-2-2-explained-new-approach-ai-video-generation) is a video generation model that produces short cinematic clips from either text prompts or static images. What sets it apart is its Mixture-of-Experts (MoE) architecture. The model switches between specialized experts that handle early scene structure and later visual refinement. This results in smoother motion and higher visual fidelity along with more efficient use of compute.

Available on Vast.ai in both text-to-video ([T2V](https://vast.ai/model/wan-2.2-t2v-a14b)) and image-to-video ([I2V](https://vast.ai/model/wan-2.2-i2v-a14b)) variants – with the latter offering optional guidance via text prompts – WAN 2.2 gives you fine control over motion, lighting, and scene composition. Popular use cases include rapid prototyping of video concepts based on text descriptions and product animations generated from still images.

> Not sure whether **WAN 2.2** or **LTX-2** fits your workflow best? Our detailed [comparison post](https://vast.ai/article/wan-2-2-vs-ltx-2-which-ai-video-model-should-you-use) explores the strengths of each model and which one to choose for your specific needs!

#### 4. ACE-Step V1 – Fast Text-to-Music Generation

[**ACE-Step V1**](https://vast.ai/model/ace-step-v1-3.5b) is an open-source music generation model that creates full musical compositions from text prompts in **17 languages**. Thanks to its diffusion-based architecture and efficient audio compression pipeline, it generates music about **15x faster** than comparable alternatives.

Musical quality is high, with controllable duration (up to five minutes works best) and coherent output across melody, harmony, and rhythm. You can even edit lyrics and manipulate vocals. Overall, ACE-Step V1 is a practical tool for fast experimentation with musical ideas, as well as voice cloning applications and music remixing and style transfer.

#### 5. Dia 1.6B – Realistic Dialogue Generation from Text

[**Dia 1.6B**](https://vast.ai/model/dia-1.6b) is a text-to-speech model that produces highly realistic dialogue in English from given transcripts. Using [S1] and [S2] speaker tags, it creates multi-speaker conversations that can also include nonverbal expressions like laughter, coughing, and gasps.

Audio conditioning enables precise control over emotion and tone, and the model even supports voice cloning functionality. Dia 1.6B's applications include conversational audio content, accessibility tools, and voice synthesis applications that require consistent, natural-sounding speech.

#### 6. DeepSeek V3.2 Exp – Long-Context Sparse Attention LLM

[**DeepSeek V3.2 Exp**](https://vast.ai/model/deepseek-v3.2-exp) is an open-source experimental large language model designed for long-context reasoning and advanced text generation. It introduces DeepSeek Sparse Attention, a novel mechanism that enables efficient processing of extended documents and conversations without sacrificing output quality.

The model is well suited for long-form document analysis and generation, research that requires extended reasoning, code generation and debugging, and even customer support with context-aware responses. Notably, it has a license that offers full commercial use permissions.

#### 7. Qwen3.5 397B A17B – Multimodal Reasoning and Agentic Workflows

[**Qwen3.5 397B A17B**](https://vast.ai/model/qwen35-397b-a17b) is the open-weight of the first model in the new Qwen3.5 series. It's a frontier-level native vision-language model with outstanding results across numerous benchmarks.

Built with a sparse mixture-of-experts (MoE) architecture and hybrid attention mechanisms, Qwen3.5 397B A17B supports highly complex tasks in areas like mathematical reasoning and problem solving, multi-turn agentic workflows, code generation and debugging, document analysis, video understanding, visual question answering, and multilingual interaction across **200+** languages and dialects.

### Launch Templates in Minutes on Vast.ai

Today's AI workloads span an incredibly wide range of modalities. The templates above represent just a small snapshot of what developers are building with on Vast.ai right now.

With the **Vast.ai** [Model Library](https://vast.ai/model-library), getting started on your own AI projects is easy. Simply browse templates by use case, select the recommended GPU hardware, and launch a production-ready environment in minutes!


Discover the most popular prebuilt templates on Vast.ai for image, video, audio, and text AI workloads — from Flux.2-Dev to Qwen3.5.

7 Popular Model Templates for AI Workloads on Vast.ai Right Now


Two of the largest corporate spend platforms in the US have independently named Vast.ai one of their fastest-growing vendors — not based on fundraising announcements or press coverage, but on what companies are actually spending money on.

**[Ramp](https://ramp.com/velocity/top-saas-vendors-on-ramp-march-2026)**, which tracks spend across 50,000+ businesses, named Vast.ai a breakout growth vendor in their March 2026 Velocity report. They categorized us under "Agent hosting" and noted that half the trending vendors on the platform are compute infrastructure providers. Their analysis:

> "AI agent workloads are moving from prototypes to production systems, and the constraint is shifting from 'which model is best' to 'where can I run this reliably.'"

Three months earlier, **[Brex](https://www.brex.com/journal/brex-benchmark-december-2025)** ranked Vast.ai **#11** in their 50 fastest-growing software vendors of 2025, based on transaction data from 35,000+ customers. [Fortune covered the report](https://fortune.com/2025/12/23/as-ai-investors-fret-over-roi-these-startups-attracted-serious-cash-customers-2025/), highlighting Vast.ai as a key infrastructure-layer company drawing real dollars alongside Groq (#12) and Supabase (#23).

That's 85,000+ businesses across two independent platforms telling the same story.

### What's driving the growth

The numbers behind these rankings reflect a shift we've been seeing on our own platform. Over the past twelve months:

- **New signups grew 27x year-over-year.** In February 2026 alone, more than 19,000 new users signed up for Vast.ai — compared to roughly 700 in February 2025.
- **Monthly active paying customers grew 16x**, from under 1,000 to over 14,000.
- **GPU supply scaled to meet demand.** Our marketplace now has nearly 20,000 GPUs across 5,000 machines from over 1,000 independent hosts — up 48% year-over-year.
- **Revenue grew 13x**, crossing seven figures per month for the first time.

This isn't growth driven by a single large customer or a one-time event. It's thousands of teams — researchers, startups, and enterprises — independently discovering that they can get GPU compute faster, cheaper, and with more flexibility on Vast.ai than anywhere else.

### Why corporate spend data matters

Fundraising rounds and press mentions are leading indicators. Corporate spend is a lagging one — it only moves after teams have evaluated, tested, and committed. When Ramp and Brex both flag the same company, it means real budgets are shifting. Not pilot programs. Not free-tier experiments. Paid production workloads.

Ramp's observation that the bottleneck is moving from "which model" to "where do I run it" is exactly the opportunity we built Vast.ai for. The platform is model-agnostic, cloud-agnostic, and contract-free. Teams deploy in seconds and pay by the second. That simplicity compounds — once a team starts, their usage tends to grow.

### What's next

We're investing in the areas that matter most to production teams: reliability, autoscaling, and deeper integrations for agentic workloads. The infrastructure layer for AI is still being built, and we plan to keep earning our position in these rankings month over month.

Thank you to the 100,000+ users who've signed up for Vast.ai, and to the 1,000+ hosts whose GPUs make this marketplace possible.

One more thing: we're also a Ramp customer ourselves. We use Ramp to manage our own corporate spend. Vendor and customer — it's a good sign when the tools you rely on are also relying on you.

---

*Links:*
- *[Ramp Velocity: Top SaaS Vendors — March 2026](https://ramp.com/velocity/top-saas-vendors-on-ramp-march-2026)*
- *[Brex: 50 Fastest-Growing Software Vendors of 2025](https://www.brex.com/journal/brex-benchmark-december-2025)*
- *[Fortune: AI Startups Attracting Serious Customer Cash in 2025](https://fortune.com/2025/12/23/as-ai-investors-fret-over-roi-these-startups-attracted-serious-cash-customers-2025/)*


Two of the largest corporate spend platforms independently ranked Vast.ai among their fastest-growing vendors — based on real transaction data from 85,000+ businesses.

Vast.ai Named Among Fastest Growing Vendors by Ramp and Brex


## Introduction

OpenAI has just released the GPT-OSS family of models, marking their return to open-weight model releases. The family includes GPT-OSS-120B (120 billion parameters) and GPT-OSS-20B (20 billion parameters), offering developers access to state-of-the-art language capabilities previously locked behind proprietary APIs.

GPT-OSS offers configurable reasoning capabilities with reasoning levels - from "low" for quick responses to "high" for complex problem-solving. The model uses OpenAI's harmony encoding system, which structures conversations and reasoning in a sophisticated format that enables fine-grained control over the model's thinking process.

This breakthrough means you can now:

-   Deploy OpenAI-quality models on your own infrastructure
-   Scale inference according to your exact needs

Vast.ai provides the perfect platform for running GPT-OSS models. With its marketplace of GPUs, you can choose the right hardware for your needs - whether running the efficient 20B model or the powerful 120B variant.

In this guide, we'll show you how to deploy both GPT-OSS models on Vast.ai using vLLM for optimized inference, with a focus on the 120B model. You'll learn how to interact with these models using the harmony encoding system for different reasoning levels.

## Setting Up the Environment

Before we can deploy our model, we need to set up our Vast.ai environment. First, install the Vast SDK:

```bash
# Install required packages
pip install --upgrade vastai
pip install --upgrade openai
pip install --upgrade openai-harmony
```

Set up your Vast API key (available from your [Account Page](https://cloud.vast.ai/account/)):

```bash
# Set your Vast.ai API key
export VAST_API_KEY="" # Your key here
vastai set api-key $VAST_API_KEY
```

## Choosing the Right Hardware

The GPT-OSS models have different hardware requirements:

-   **GPT-OSS-120B**: Requires an H100 GPU with 80GB VRAM
-   **GPT-OSS-20B**: Requires 16GB+ VRAM

For this guide, we'll demonstrate with the 120B model on H100. Let's search for suitable instances:

```bash
# Search for suitable GPU instances
vastai search offers " \
gpu_name = H100_SXM \
geolocation=US \
static_ip = true \
direct_port_count >= 1 \
verified = true \
disk_space >= 120 \
rentable = true"
```

## Deploying GPT-OSS with vLLM

Now let's deploy our GPT-OSS instance using vLLM's optimized inference server. We'll use the latest vLLM image with special GPT-OSS support. The pip install commands are from the model's Hugging Face pages ([120B](https://huggingface.co/openai/gpt-oss-120b), [20B](https://huggingface.co/openai/gpt-oss-20b)).

### Deploying GPT-OSS-120B (H100 recommended)

```bash
# Deploy vLLM instance for 120B model
export INSTANCE_ID= # Insert instance ID from search results

vastai create instance $INSTANCE_ID --image vllm/vllm-openai:latest --env '-p 8000:8000' --disk 160  --onstart-cmd 'uv pip install --system --upgrade transformers kernels torch openai; uv pip install --system --pre vllm==0.10.1+gptoss --extra-index-url https://wheels.vllm.ai/gpt-oss/ --extra-index-url https://download.pytorch.org/whl/nightly/cu128; vllm serve openai/gpt-oss-120b --max-model-len 80000'
```

### Alternative: Deploying GPT-OSS-20B on A100

Here's how to deploy the 20B model on an A100 GPU:

```bash
# Deploy vLLM instance for 20B model
export INSTANCE_ID= # Insert instance ID from search results

vastai create instance $INSTANCE_ID --image vllm/vllm-openai:latest --env '-p 8000:8000' --disk 80  --onstart-cmd 'uv pip install --system --upgrade transformers kernels torch openai; uv pip install --system --pre vllm==0.10.1+gptoss --extra-index-url https://wheels.vllm.ai/gpt-oss/ --extra-index-url https://download.pytorch.org/whl/nightly/cu128; vllm serve openai/gpt-oss-20b'
```

After deployment, wait for the model to download and start serving. You can monitor the logs to see when it's ready. Once running, find your instance's IP address and port from the [Instances tab](https://cloud.vast.ai/instances/) in the Vast AI Console.

## Interacting with GPT-OSS Using Harmony Encoding

GPT-OSS uses OpenAI's harmony encoding system, which provides structured conversation formatting and enables different reasoning levels. Let's set up our client:

```python
from openai import OpenAI
from openai_harmony import (
    load_harmony_encoding,
    HarmonyEncodingName,
    Role,
    Message,
    Conversation,
    SystemContent,
    DeveloperContent,
)

# Your server details
VAST_IP_ADDRESS = ""
VAST_PORT = ""

# Initialize client
client = OpenAI(
    api_key="EMPTY",
    base_url=f"http://{VAST_IP_ADDRESS}:{VAST_PORT}/v1"
)

# Load harmony encoding
enc = load_harmony_encoding(HarmonyEncodingName.HARMONY_GPT_OSS)

## Simple Chat Function
def chat_gpt_oss(prompt, reasoning="medium"):
    """
    Simple chat with GPT-OSS
    reasoning: "low", "medium", or "high"
    """
    # Create conversation with harmony format
    conversation = Conversation.from_messages([
        Message.from_role_and_content(Role.SYSTEM, SystemContent.new()),
        Message.from_role_and_content(
            Role.DEVELOPER,
            DeveloperContent.new().with_instructions(f"Reasoning: {reasoning}")
        ),
        Message.from_role_and_content(Role.USER, prompt),
    ])

    # Get formatted prompt
    formatted_prompt = enc.render_conversation_for_completion(conversation, Role.ASSISTANT)

    # Make API call
    response = client.completions.create(
        model="openai/gpt-oss-120b",
        prompt=formatted_prompt,
        max_tokens=2000,
        temperature=0.7,
    )

    return response.choices[0].text


prompt = "Explain quantum computing to a 4th grader. Keep in concise."
```

## Comparing Reasoning Levels

One of GPT-OSS's most interesting features is its configurable reasoning levels. Let's see how the model responds differently to the same prompt with low and high reasoning:

### Low Reasoning Response

```python
result = chat_gpt_oss(prompt, "low")
print("----LOW REASONING----")
print(result)
```

Output:

```
----LOW REASONING----
analysisWe need to explain quantum computing to a 4th grader, concise. Use simple analogies, like bits as light switches, qubits as coins that can be heads, tails, or both. Keep short.assistantfinal**Quantum Computing in a Nutshell (for a 4th‑grader)**

1. **Regular computers = light switches**
   - A normal computer works with bits that are either **0** (off) or **1** (on), just like a light switch that's either down or up.

2. **Quantum computers = magic coins**
   - A quantum computer uses **qubits**.
   - Think of a qubit like a coin that can be **heads (0)**, **tails (1)**, **or both at the same time** while it's spinning in the air. That "both at once" is called **superposition**.

3. **Working together**
   - When you have many regular bits, you have to try every possible combination one after another.
   - With many qubits, they can be in many combinations **all at once** because they can be linked together in a special way called **entanglement**. It's like a group of magic coins that always know what the others are doing, even if they're far apart.

4. **Why it matters**
   - Because of superposition and entanglement, a quantum computer can solve some puzzles (like cracking codes or simulating molecules) **much faster** than a regular computer.

**Bottom line:** A regular computer is like a row of on/off switches; a quantum computer is like a bunch of spinning magic coins that can be many things at once, letting it try many solutions at the same time.
```

### High Reasoning Response

```python
result = chat_gpt_oss(prompt, "high")
print("----HIGH REASONING----")
print(result)
```

Output:

```
----HIGH REASONING----
analysisWe need to explain quantum computing to a 4th grader, concise. Use simple analogies, avoid jargon, but keep accurate. Should be short. Provide analogies like coins, dice, superposition, entanglement. Keep concise.assistantfinal**What is a quantum computer?**

Imagine you have a super‑smart robot that can solve puzzles much faster than a regular computer. A quantum computer is that robot, but instead of using normal "on/off" switches (bits) like a regular computer, it uses tiny, magical switches called **qubits**.

### Bits vs. Qubits
- **Bit (regular computer):** Like a tiny light that can be **off (0)** or **on (1)**.
- **Qubit (quantum computer):** Like a tiny spinning top that can be **0**, **1**, **or both at the same time** (this is called *superposition*).

### Why is that cool?
If you have 10 bits, they can show only one of 2¹⁰ = 1,024 possible patterns at a time.
If you have 10 qubits, they can be in all 2¹⁰ = 1,024 patterns **at once**! That lets a quantum computer try many solutions together.

### A Simple Analogy
Think of a maze puzzle:

- **Regular computer:** Sends a single mouse down the maze, step by step, until it finds the exit.
- **Quantum computer:** Sends a whole swarm of mice that can explore every possible path at the same time and tell you the exit almost instantly.

### Entanglement (Teamwork)
Sometimes two qubits become "best friends" and link together, no matter how far apart they are. Changing one instantly changes the other. This is called **entanglement**, and it lets quantum computers coordinate their work in a super‑efficient way.

### Bottom Line
A quantum computer uses qubits that can be 0, 1, or both at once, and they can be entangled with each other. This lets the machine look at many possibilities at the same time, solving certain problems way faster than ordinary computers.

So, it's like a super‑fast, super‑clever puzzle‑solver that can try lots of answers all at once!
```

## Comparing Low vs High Reasoning Outputs

The quantum computing explanations demonstrate the distinct capabilities of GPT-OSS-120B's reasoning levels. With low reasoning, the model produced a concise, structured explanation using simple analogies like "magic coins" and "light switches" - delivering clear concepts in a compact format suitable for quick understanding.

The high reasoning response revealed significantly more sophisticated thinking. The model provided detailed analogies like the maze puzzle with mice, explained mathematical concepts (2¹⁰ patterns), and structured the explanation with clear sections and formatting. Most notably, it offered multiple perspectives on the same concepts - from the spinning top analogy to the teamwork explanation of entanglement.

This difference in reasoning depth represents a significant advancement in model control. Rather than simply adjusting temperature or other parameters, GPT-OSS allows direct specification of how thoroughly the model should think through problems. The low reasoning approach prioritized efficiency while maintaining accuracy, while high reasoning prioritized comprehensive understanding with rich examples and detailed explanations.

## Conclusion

OpenAI's first open-source model thus introduces not just accessibility to advanced language capabilities, but a new paradigm for controlling model reasoning depth - enabling applications to dynamically adjust between quick responses and thorough analysis based on context and requirements.


Learn how to deploy and run OpenAI's GPT-OSS models on Vast.ai using vLLM, with configurable reasoning levels and harmony encoding system.

Running OpenAI's GPT-OSS on Vast.ai

Deploy Llama 4, Qwen3, and Kimi K2 on Vast.ai in minutes. Use optimized templates like Oobabooga, vLLM, HuggingFace TGI, and Open WebUI—no setup hassle.

Modular MAX vs vLLM Performance Comparison on Vast.ai


Mukesh Ambani's Reliance Group is [making waves](https://www.bloomberg.com/news/articles/2025-01-23/billionaire-mukesh-ambani-plans-world-s-biggest-data-center-in-india-s-gujarat) with its plans to build what may become **the** **world's largest data center by capacity** in Jamnagar, India. The facility, expected to reach three gigawatts, would dwarf today's largest data centers, which are under one gigawatt each (or will just barely exceed one gigawatt once fully operational, as in the case of the Sines data center in Europe).

Reliance isn't alone in its bid to drastically expand AI infrastructure – companies around the world are racing to meet surging demand in this sector. Tech giants like Microsoft, Alphabet, and Amazon are investing billions of dollars into data centers to expand their AI capabilities.

The Stargate Project is another entity endeavoring to boost AI infrastructure in the United States. OpenAI, SoftBank, and Oracle all recently [pledged](https://techcrunch.com/2025/01/21/openai-teams-up-with-softbank-and-oracle-on-50b-data-center-project/) to invest $100 billion to $500 billion as part of the Stargate effort.

So what makes Reliance interesting? The company has some big ambitions...

### Reliance's AI Strategy

Ambani, the 67-year-old billionaire chairman and managing director of Reliance Group, hopes to offer AI inference to businesses and developers at significantly lower costs than competitors. In order to achieve this goal, Reliance is investing heavily in NVIDIA's AI semiconductors.

Undercutting prices isn't a new approach for Ambani; his aggressive pricing in other industries, like telecommunications, disrupted the market and forced competitors to adapt. It's not unreasonable to think that Reliance could do the same in the AI industry.

"By leveraging our expertise in infrastructure, networking, operations, software, and data and by collaborating with our global partners, our goal is to create the world's lowest AI inferencing cost, right here in India," Ambani [said](https://www.bloomberg.com/news/articles/2025-01-23/billionaire-mukesh-ambani-plans-world-s-biggest-data-center-in-india-s-gujarat) last year. "This will make AI applications in India more affordable than anywhere else, making AI accessible to all."

### Strengthening India's AI Infrastructure

One of Reliance's main objectives with this massive data center is to reduce India's reliance on overseas AI infrastructure. As AI adoption accelerates across the country, driven by a booming startup ecosystem and increasing enterprise demand, the need for localized AI compute power is greater than ever.

Currently, India's total data center capacity is under one gigawatt. If completed as planned, the Jamnagar facility alone could triple that amount by 2027\. Reliance also reportedly aims to power the new data center as much as possible with renewable energy – aligning with global sustainability trends and potentially positioning India as a more self-sufficient AI hub.

Beyond meeting domestic demand, this expansion raises questions about India's role in the global AI race. With governments and tech giants worldwide investing heavily in AI infrastructure, Reliance's project reflects India's potential to compete on the world stage as a provider of AI compute power.

### What's Next for AI Infrastructure?

Reliance's plan for the facility in Jamnagar is just one of many data center projects in the works today. According to McKinsey & Co, global data center capacity could [more than triple](https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/ai-power-expanding-data-center-capacity-to-meet-growing-demand) by 2030, reaching an estimated 219 gigawatts annually. Yet even with heavy investment, a significant shortfall will likely remain. **The U.S. alone is projected to face a 15-gigawatt gap by 2030**, requiring twice the data center capacity that's been built since 2000\.

If Reliance successfully builds its three-gigawatt facility, it could help address some of this global demand. However, ensuring that infrastructure keeps pace with AI's rapid advancement will remain a challenge worldwide.

In the meantime, access to high-performance compute remains a priority for developers, researchers, and businesses. While tech giants build massive facilities, **cloud GPU rental is an option that provides immediate, scalable access to AI compute** without the need for expensive hardware or infrastructure investments.

If you're looking for flexible and cost-effective AI compute, Vast.ai offers on-demand cloud GPU rental. We invite you to explore our platform and get the AI compute you need – and start scaling your workloads today\!


Reliance plans to build the world’s largest data center in India, aiming to lower AI costs and boost local infrastructure with NVIDIA chips and renewable energy.

Posts about: AI Infrastructure

7 Popular Model Templates for AI Workloads on Vast.ai Right Now

Vast.ai Named Among Fastest Growing Vendors by Ramp and Brex

Running OpenAI's GPT-OSS on Vast.ai

Modular MAX vs vLLM Performance Comparison on Vast.ai

India's AI Leap: World's Biggest Data Center on the Horizon?