Baseten

What do you like best about the product?

It’s easy to deploy LLMs like Llama, Mistral, etc. on dedicated GPUs, and the API endpoint is clean and straightforward. I also like that I don’t need to manually manage a complex Kubernetes cluster to get everything running.

What do you dislike about the product?

Pricing doesn’t seem to scale well for heavy, enterprise-level GPU workloads, and I also wish the documentation were more granular—especially around custom Triton server setups.

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

It helps reduce the infrastructure overhead of self-hosting machine learning models while also lowering inference latency for production AI applications.

What do you like best about the product?

Packing and shipping models with Truss is remarkably straightforward, especially compared with managing raw Kubernetes clusters or maintaining custom Triton setups.

What do you dislike about the product?

Debugging cryptic model packaging errors or dependency conflicts during container builds can take some trial and error in the CLI.

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

Our team needed dedicated, high-throughput GPU infrastructure to run production ML models, without having to build an internal MLOps team from scratch.

What do you like best about the product?

I really appreciate how simple Truss makes deployment; we were able to move from a local model to a production endpoint in just a few hours thanks to the accurate documentation and examples. I also like how Baseten's autoscaling and optimized inference provided us with reliable, low-latency performance without overpaying for idle GPUs. Switching to Baseten from self-hosted GPUs and SageMaker offered us faster deployment and lower latency. Plus, not having to write Dockerfiles, handling dependencies, and model caching easily with Truss makes it even more useful for me.

What do you dislike about the product?

There's not really much that I have issues with, but if I had to pick something, it would be that some of the more advanced Truss configurations could use better documentation. Support helped us out quickly when we needed it, but it would be nice to have more end-to-end examples for things like private model weights with secrets and performance tuning, like concurrency and batching settings. The basic examples are great; it's just those edge cases that needed a bit more guidance.

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

I use Baseten to deploy fine-tuned models, reducing latency and costs with its autoscaling. Truss simplifies deployment without needing Dockerfiles, supporting local testing and dependency handling.

What do you like best about the product?

Mainly, I’ve used baseten to run and deploy ML models for testing and production deployment.

There are a lot of things I really like about it, including the good UX, the speed, support documentation and is very cheap comparing other platforms.

What do you dislike about the product?

It doesn’t have strong downstream integration with other apps we use in our workflows. Also, the AI features feel quite limited compared with other platforms.

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

It really helps simplify our deployment process and reduces the infrastructure work on our side. It also makes it easier to use the deployed models through APIs.

Overall, this helps us build them more easily and then use them reliably later on.

What do you like best about the product?

I really like how easy it is to deploy open-source models like DeepSeek. The dashboard is very clean and simple to use. Setting up my first model API took less than ten minutes, which saved me a lot of time

What do you dislike about the product?

The pricing can be a bit confusing for new users who are just starting out. Also, I noticed that the documentation for advanced model configurations is slightly hard to follow if You Don't Have a strong coding background

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

It saves a lot of time because I do not have to worry about managing heavy cloud infrastructure or servers manually. I can just focus on testing different AI models quickly, which makes my development workflow much smoother

What do you like best about the product?

I really appreciate Baseten because it allows for production-grade inference without the headache. I use it for high-performance AI inference and model deployment, and it solves the problem of needing production-grade models without GPUs by providing dedicated managed GPU infrastructure. The initial setup was smooth and surprisingly fast, which was a big plus for me.

What do you dislike about the product?

If Baseten reduces the cost overhead of dedicated replicas, it would really help us improve. It would be great if they could introduce on-demand replica activation without a cold start.

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

I use Baseten for production-grade AI inference, benefiting from smooth setup and managed GPU infrastructure without headaches.

What do you like best about the product?

Baseten’s superpower is turning a custom or open-source model into a production API without me becoming a full-time GPU SRE. Truss packaging, brutal cold-start work, autoscaling with scale-to-zero, and multi-cloud reliability mean I get low latency and high throughput without babysitting replicas. Model APIs when I want speed-to-first-call, dedicated deployments when I want control — same stack, SOC 2/HIPAA when I need it. That’s what I like most.

What do you dislike about the product?

Pricing. On-demand GPUs cost more than raw clouds, billing is per minute not per second, and if you keep a replica warm for latency you pay for idle time

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

The benefit is time and reliability. I ship inference instead of babysitting infra, latency stays usable under load, and I’m not paying for idle GPUs 24/7. Multi-cloud capacity plus SOC 2/HIPAA also means I can run serious workloads without building a second platform team

What do you like best about the product?

Deployment workflow with Truss is easily the strongest feature. Packaging models with their specific dependencies, CUDA versions, and environment configurations used to be a major bottleneck for our engineering pipeline. Baseten abstracts almost all of that containerization headache away. Autoscaling works as advertised under bursty traffic, and cold starts on GPUs are noticeably quicker than configuring custom Kubernetes clusters on raw cloud providers. The observability dashboard gives clear real-time visibility into latency, throughput, and error rates right out of the box...

What do you dislike about the product?

It's debugging failed deployments or subtle dependency issues during build time can sometimes feel opaque, often requiring local reproduction through the CLI to catch where an environment configuration failed. Additionally, dedicated GPU resource pricing can escalate quickly if autoscaling thresholds and scale-to-zero settings are not closely monitored and configured properly from day one!...

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

Baseten solved our machine learning deployment lifecycle. Prior to using Baseten, shipping custom LLMs and fine tuned open-source models required dedicated MLOps support to manage GPU clusters, optimize inference runtimes, and maintain API uptime. Baseten allows our developers to go from a model checkpoint to a production-ready API endpoint in minutes instead of days, saving dozens of engineering hours each sprint...

What do you like best about the product?

I like how Baseten allows me to deploy and run AI models in production without needing to build all the infrastructure myself. It's really convenient for managing inference. Deployment and scaling are much easier, so I can focus on the models themselves. I really appreciate the model deployment tools and autoscaling, as they make it simpler to get models into production and handle traffic changes without manual intervention. Also, Baseten fits well into our existing AI workflow with our current cloud and development tools.

What do you dislike about the product?

The learning curve can be a bit steep at first, especially when setting up more advanced deployments. Better documentation or similar setups for beginners would help. The advanced development setting can be confusing at first. More step-by-step guides, practical examples, and beginner-friendly tutorials would make the setup easier.

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

I use Baseten to deploy and run AI models in production, saving me from infrastructure work. It simplifies deployment and scaling, letting me focus on the models. I value the model deployment tools and autoscaling for handling traffic changes without manual management.

What do you like best about the product?

What I like best about Baseten is how easy it makes deploying and scaling AI models in production. The combination of fast inference, autoscaling, and good developer tooling means I can focus more on the model and application rather than managing GPU infrastructure. I also like the flexibility to deploy custom or fine-tuned models while still getting strong performance and observability.

What do you dislike about the product?

What I dislike about Baseten is that it can take a while to get comfortable with the platform, particularly when you’re setting up more advanced deployments. Pricing can also be hard to predict if your workloads have variable usage, and some features can feel better suited to teams with strong technical expertise.

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

Baseten helps cut through the complexity of deploying, scaling, and managing AI models in production. It reduces the time I have to spend on infrastructure work, GPU resources, and performance optimization. As a result, I can get AI applications into production faster, keep performance more reliable over time, and focus more on improving the product itself instead of managing the underlying infrastructure.

Recommendations to others considering the product:

To improve Baseten, I would suggest enhancing the onboarding process to make it more intuitive for new users. Additionally, providing clearer pricing models and offering more resources for users with varying levels of technical expertise could be beneficial.