There are several rundowns between the two services like this one.

Key for me:

• Only Kaggle supports R, only Colab supports SWIFT
• Colab is a Google product and is therefore optimized for Tensorflow over Pytorch
• Colab is a bit faster and has more execution time (9h vs 12h)
• Yes Colab has Drive integration but with a horrid interface, forcing you to sign on every notebook restart

Kaggle has a better UI and is simpler to use but Colab is faster and offers more time.

I am new to both Kaggle and Colab. But in these days I have found some advantages of Kaggle over Colab.

1. Kaggle is a great source of datasets and you can use these Datasets in Kaggle notebooks very easily. There is no need to download them, just add them. In Colab, you need to download the dataset. However, if you want to use very own datasets, then you need to upload it for the first time.

2. Kaggle provides 35 hours GPU usage per user in a week and also show you how much time you have used. Colab has no such mentions but they also limit usage of GPU and they won't say how much time you have used and how much time it will be available. This sometimes leads to problem in deciding when to use GPU and when not to.

3. Google Colab notebooks need to be open and active during the using and training time, while you can commit a kaggle notebook then close it if you want to come later and see the training results.

You can find more comparisons here.

Notes: 1) Kaggle now provides StackOverflow instance search. 2) Both Kaggle and Colab are now Google subsudiary.

I have been frequently using both the platforms and could straightaway point a few differences:

1. Kaggle gives NVIDIA Tesla P100 PCI based 16GB GPUS for approximately 9 straight hrs in a single commit, whereas Colab provides NVIDIA Tesla K80 GPU 12 GB for 12hrs.
2. Kaggle has a limitation of 5 GB hard-drive space vs Colab's storage could vary from 30GB to 72GB as per the availability.
3. Kaggle provides TPU v3-8 with whereas Colab has not disclosed its TPU version anywhere.
4. I almost never required to install any third-party package for any of my data science work, whereas in Colab it gets frustrating sometimes (Broken dependencies and package incompatibilities).

Kaggle cons:

1. Weekly limit to GPU and TPU usage. (Although this limit is almost sufficient for basic training)
2. Limited storage (If you go above 5GB, you will face a kernel crash)

Colab cons:

1. Not consistent in performance as it changes hardware resources as per the availability in the pool.

I have also faced issues where if you use your notebook kernel for above 12hrs, Colab reduces the hardware grant in further commits and degrades the performance. The only best thing in Colab is the drive mounting facility which is very handy at times. Kaggle on the other hand if very handy if you adapt to the interface and its API.