The Airtrain Playground: Match up Mistral, Gemini, GPT-4, Phi-2, Llama 2 and more

At Airtrain AI, our goal is to facilitate your transition from costly AI APIs such as GPT-4 to small inexpensive models customized for your application.

The first step towards moving away from proprietary AI models is to evaluate the performance of alternatives. Last year we launched our batch evaluation product to run evaluation tasks on large datasets. Today we are augmenting our suite of LLM-focused tools with the LLM Playground.

With Airtrain's LLM Playground, you can chat and interact with a large selection of open-source and proprietary models. Prompt once and get all selected models to respond at once. Then compare results and iterate until you find a suitable model for your application. Then you can move on to our batch evaluation product to evaluate models at scale.

At this time, the Airtrain Playground supports the following models:

• OpenAI: GPT-3.5 Turbo†, GPT-4
• Mistral AI: Mistral 7B, Mixtral 8x7B, Mistral Medium
• Google: Gemini Nano, Gemini Pro, FLAN-T5 XL, XXL
• Microsoft: Phi-2
• Llama 2 7B, 13B, 70B
• Falcon 7B

The Airtrain Playground is free to use, simply sign up and click "Play with Models".

If you need assistance getting started or obtaining third-party API tokens for proprietary models (GPTs, Gemini, etc.), join our Slack to get help or give feedback!