Mistral Small 4 Free API

Key Takeaways

So, hi everyone. Welcome back to Data Science in a pocket and Mistral has released a big big model, but the name is small. That is Mistral Small 4 with 119 billion parameters. Introducing Mistral Small 4 as you can see. This is one of the best models that the team has released so far in the Mistral Small family. A new standard for multi-model reasoning optimized model. As you can see here, what are the key details of the architecture? Mixture of expert as usual. 128 experts, 119 total parameters with 6 billion active parameters. The contest window is quite interesting, 256K. Configurable reasoning effort. I think this has something new that the team has introduced and native multi-modality. Accept both text and image inputs. What is the performance? 40% reduction in end-to-end completion time. Three cross more requests per second. That is very interesting. If you look at the chart, instruct and reasoning models, you can see that the model is quite good. Why Mistral Small 4 unified capabilities? I think it's it's consolidate the strengths of Magistral, Destral, and Mistral Small. All the three models combined into one. An interesting part they have released is reasoning on demand. Reasoning effort is none, fast. Reasoning effort is like deep. So, the same model can be used without reasoning also and with reasoning also. That is a very important button that the team has given. These are the benchmarks and the model is quite good, quite interesting as you can see. But, we are not here to discuss the benchmarks of the model, but how to use the model for free using the free API key. Because to run 119 billion parameters in local system is just not possible. So, here we are. As you can see, we are NVIDIA's Jim's page where the model is available to use for free. Yes, you heard it right. Mistral Small 4 119 billion parameters. The model can be tested out for free here. Let me hit this particular prompt. Is there a car in this image? I will just hit enter and you'll be able to use a model. Yes, there's a car in the image. You are able to use the model for free inferencing and you get the free API also. So, you just need to swap to view code section and in the view code section, you can see that there's a Python option there where you get the entire code here. Here's the API key. To generate the API key, you need to click on generate API key here. You'll be able to get your free API key now. Once you are able to get your free API key, just copy the entire code. Click on the top left. So, here you can see that what are the API rate limits? Up to 40 RPMs per minute. That is quite huge. I think for testing purpose, it's pretty good. Now, what is the next step? I copied the entire code here in Mistral Small 4 in this Google Colab notebook. I ran the code and with ease, you are able to get the output here. Hey buddy, how's it going? It's able to give an answer here. I asked it a question. Now, let me ask something else. What is life? So, let's hit enter again. Here also, you'll be able to get an answer. Life has an explored is a profound and multifaceted concept that has been explored in philosophy, etc., etc. So, one of the best models from Mistral team that is Mistral Small 4 is released now. The API key is also free on NVIDIA NIMS page. Yes, I have shared the link in the comment section below and you can use it for free with the rate of 40 RPM. So, this is quite decent for experimental purposes. I will you try out the new model. Looks quite good. Thank you so much.