⚙️ Playing with Models: Groq, OpenRouter and Others #MLSTART

When you start understanding how LLMs work, you immediately want to try everything in action - "feel" the models with your own hands, compare generation quality, see how certain parameters affect the LLM response.

Now there are plenty of convenient platforms and APIs for this, where you can test different models without setting up infrastructure.

Here are a few services 👇

- Groq — a cool service, but requires a VPN. You can run open-source models for free, the limits are enough to comfortably play around and test ideas.

- OpenRouter - an alternative without VPN. There are also free models there (limits are slightly smaller), but the interface is convenient, and you can quickly compare different LLMs.

- Replicate and Together.ai - more about production and experiments. These services are paid, require a foreign card, but they have a huge "zoo" of models, not only LLMs: there are models for images, audio, and video.

💡 Such tools are useful to quickly test a hypothesis, understand how a model reacts to different prompts, and not waste time on deploying a local environment.

If you want to quickly test different LLM models - these playground platforms are a great place to start.

Back to table of contents

👩‍💻 Data Flow