Friendli Suite

What do you like best about the product?

What I like best about Friendli AI is its focus on fast and efficient AI model inference, helping applications generate responses with low latency and optimized resource utilization. It supports scalable AI deployment and makes it easier to integrate AI capabilities into business workflows, improving automation potential, productivity, and overall operational efficiency.

What do you dislike about the product?

One area for improvement is making model deployment and configuration more intuitive, especially for users without a strong technical background. Clearer documentation, simpler integration options, and more transparent pricing and performance comparisons would make Friendli AI easier to adopt and manage for different business needs.

What problems is the product solving and how is that benefiting you?

Friendli AI helps optimize AI model deployment and inference, enabling faster response generation and more efficient use of computing resources. This can reduce latency, support scalable AI-powered workflows, and improve the efficiency of automation initiatives. It helps businesses explore AI solutions that reduce manual effort, streamline operations, and enhance overall productivity.

What do you like best about the product?

I really like the inference speed they provide. We mainly fine-tune several models for our document extraction, so inference needs to be fast, and this platform has been reliable, quick, and works well for us. The dashboard also has a very good user interface, and the support documentation helped us get set up without spending a lot of time at the beginning. Integration with other platforms, mainly Hugging Face, was smooth as well. I also really like their AI-based inference optimizations.

What do you dislike about the product?

Their pricing model isn’t very transparent, since it’s charged per endpoint and per GPU hour, so the overall cost can add up quickly. Also, the advanced settings for modifying our AI models took some time to understand and figure out.

What problems is the product solving and how is that benefiting you?

The main challenge for us was finding a reliable, fast platform where we could host our AI models. Previously, we ran our models on other platforms, but they came with several issues. With this platform, any modification is quick, and it’s reliable, which helps ensure our main pipeline runs fast and consistently. It has also improved our processing time overall.

What do you like best about the product?

We can deploy Friendli AI with our own hardware, containers or serverless APIs. This creates flexibility for where we choose to use the product and keep the deployment as close as possible to the use case. Onboarding is made easier since we don't have to buy brand new hardware.

What do you dislike about the product?

The UI can use a bit more polish. It is getting there but I think as the product continues to mature, the UX will come along with it.

What problems is the product solving and how is that benefiting you?

Friendli is helping us with integrations throughout the entire organization. We have a lot of data points and this is really helping to tie it all together.

What do you like best about the product?

What I appreciate most is its focus on high-performance AI inference without the operational complexity of managing GPU infrastructure. The combination of low latency, high throughput, and autoscaling is particularly valuable. It has OpenAI-compatible APIs and built-in monitoring, making it easier for us to integrate AI workloads into existing flows while maintaining scalability, reliability, and cost control.

What do you dislike about the product?

The main drawback is limited integration with enterprise observability platforms and, at some scale, a complex onboarding process.

What problems is the product solving and how is that benefiting you?

Complexity and cost of deploying and scaling AI inference workloads.

What do you like best about the product?

I appreciate that Friendli AI makes working with AI models feel more manageable, especially while I’m still learning the technical side. It makes it easier to test different models without worrying too much about the infrastructure, and helps me understand their speed and responsiveness. It saves me time on the technical side so I can focus more on exploring what AI can do and how I can apply it to my everyday tasks. The initial setup was straightforward.

What do you dislike about the product?

I think Friendli AI could be a little more approachable for users without a strong technical background. Some features and settings take a bit of time to understand, especially when I’m just experimenting rather than developing AI applications professionally.

What problems is the product solving and how is that benefiting you?

I can run and test different AI models without getting caught up in complicated infrastructure or technical setup. Friendli AI saves me time on the technical side and helps me understand model speed and responsiveness, so I can focus on exploring what AI can do and how to apply it to my everyday tasks.

What do you like best about the product?

I like being able to serve Hugging Face models, even one I fine-tuned, without having to set up infrastructure. Friendli AI reduces the process to an HTTP call and pay-as-you-go, and the implementation was super fast: I just had to change the URL.

What do you dislike about the product?

I wish there was a permanent free plan; that could improve.

What problems is the product solving and how is that benefiting you?

By running open models, I don't have to manage GPUs, drivers, or scaling: Friendli AI reduces it to an HTTP call and pay for what I use.

What do you like best about the product?

I like Friendli AI’s fast inference and how easy it is to deploy models. It helps us get models up and running quickly, deliver low-latency responses, and handle multiple requests and growing workloads more efficiently. The straightforward deployment process cuts down on manual infrastructure configuration, troubleshooting, and the time we spend managing model serving, so we can focus more on developing, testing, and improving AI applications. The documentation and deployment workflows made it relatively easy to get started, and integrating it into our existing workflow was manageable. It also helps us keep infrastructure costs under control.

What do you dislike about the product?

I found the learning curve a bit challenging for users new to AI model deployment and inference optimization. Some advanced settings could be explained more clearly with documentation and practical examples, and understanding advanced settings and optimizing models took me extra time. I’d also like more detailed usage analytics and cost breakdowns, clearer logs, real-time performance metrics, and alerts for unexpected increases in latency or resource use. Better integration with popular ML apps and observability tools, along with more transparent pricing estimates and cost predictions before deployment, would also help.

What problems is the product solving and how is that benefiting you?

I use Friendli AI to serve models efficiently, improve response times, and handle increasing workloads without spending too much time managing infrastructure. It addresses high infrastructure costs and the complexity of deploying and scaling AI models.

What do you like best about the product?

I use Friendli AI whenever I want to run and test open-source language models without having to manage heavy server setups. The response speed and streaming outputs are noticeably fast compared to regular hosting platforms. Connecting the API into my projects is simple because the setup is very standard and works right away. It makes testing different model sizes straightforward and saves me from having to configure everything manually just to see how a model performs.

What do you dislike about the product?

Sometimes it can be a little hard to compare token usage, pricing, and latency across different models directly inside the dashboard. You have to spend some extra time checking the docs or switching between pages to figure out the exact cost differences. It would also be nice to have a few more basic starter examples or templates in the documentation to make testing new endpoints even quicker.

What problems is the product solving and how is that benefiting you?

Setting up and hosting open-source models on your own cloud servers usually takes a lot of time, expensive GPU rentals, and constant maintenance. Friendli AI solves this by giving you fast, ready-to-use endpoints on a pay-as-you-go setup. This helps me experiment with AI features and build apps much faster without having to deal with backend server headaches or high fixed monthly costs.

What do you like best about the product?

I like Friendli AI's user experience and answers, and I enjoy the experience. The initial setup was very easy.

What do you dislike about the product?

I think Friendli AI should focus on customizing messages to users based on their responses. Getting answers could be better.

What problems is the product solving and how is that benefiting you?

n/a

What do you like best about the product?

What stands out most to me is the impressive inference speed and consistently low latency. Model setup is quick and straightforward, and the serving engine can handle high concurrency without sacrificing performance. Overall, it makes running open-source LLMs in production faster and more cost-effective.

What do you dislike about the product?

Overall, there isn’t much to dislike, but I do wish the documentation and advanced customization guides went a bit deeper—especially for niche edge cases. A few additional step-by-step tutorials for specific custom model setups would reduce the trial and error during initial integration and make the onboarding process smoother.

What problems is the product solving and how is that benefiting you?

Serving open-source generative AI models at scale can get expensive very quickly because of GPU requirements. Friendli AI addresses this by maximizing throughput and optimizing memory usage, which significantly reduces infrastructure costs. For us, that translates into consistently fast response times for users without blowing through our hardware budget.