DeepSeek-R1 is now available in GitHub Models (Public Preview)

[Akash1134]

The latest trending AI model DeepSeek-R1 is now available in GitHub Models.

DeepSeek-R1 is a 671B parameter AI model designed to enhance deep learning, natural language processing, and computer vision capabilities. It offers a wide range of possibilities, provides quick insights, and allows users to explore the potential of AI in various applications.

GitHub Models makes it easy for every developer to build AI features and products on GitHub.

Try, compare, and implement this model in your code for free in the playground or via the API. Compare it to other models using side-by-side comparisons in GitHub Models.

To learn more about GitHub Models, check out the docs.

Got feedback or questions? Please share them in the comments. 🚀

[leonsparrowJM]

Why would you want to send all of our IP to the Chinese government?

[llightcb]

i am sure you have forgotten that the U.S. has... changed? they are no longer in a position to talk about other 'evil governments.' insanity, wherever i look. heil musk. not.

[LindseyB]

The model is hosted on Azure. So it's not hosted internationally if you are worried about your data.

[MathiasSchindler]

Dear @Akash1134 , thanks for this information. I did give it a try in the playground yesterday and after entering the prompt, I saw a long train of thought over two minutes until the output was replaced by an error message ("An error occurred. Please try again"). The prompt was not sensitive in any way relevant to the model. Could this have been some sort of timeout?

[Eneswunbeaten]

I think this is because the R1 thinks inside itself between <think></think> tags. Most of the new generation models do this. If Github integrates this, we won't get this error in the thinking layer.

[LindseyB]

We're still stabilizing the model and R1 takes a bit longer to respond than we expected so we increased some timeouts and it should happen a bit less often now (if at all 🤞🏻).

[am-abudu]

I noticed that the rate limit tier of DeepSeek-R1 is High, but when i try to use it, it tells me that the rate limit is 1per min.
Are there any mistake?
Error Message:

[openai_api_compatible] Error: API request failed with status code 429: {"error":{"code":"RateLimitReached","message":"Rate limit of 1 per 60s exceeded for UserByModelByMinute. Please wait 37 seconds before retrying.","details":"Rate limit of 1 per 60s exceeded for UserByModelByMinute. Please wait 37 seconds before retrying."}}

[LindseyB]

edited for a correction: Our documentation is wrong, but our actual rate limit is correct. The rate limit tier for DeepSeek is free not high we will be updating our documentation to reflect that

[jpzwarte]

Just tried this:

The error suddenly appeared after it had responded with a full page of text.

[LindseyB]

Yeah, frustratingly due to us having some stabilization issues, we've worked out some of the kinks and it should be working a bit better now. Give it another try.

[thedjswarm]

It is soooo slllooowww... and timed out at 5 min. Not usable. :(

[LindseyB]

We've done some work to stabilize a little bit, but yes the model is a bit slower than expected want to give it another try?

[barrosohub]

Not working! "Rate limited"

[juan-jurado]

Hey! Did you find any rate limit number?

[barrosohub]

Solved.

Apparently it was a general and generic error! Github itself reported that in the first hours of launch and availability within Github Models, the model was intermittently available.

[LindseyB]

Rate limit is Rate limit of 8 per 86400s for paid users

[eevmanu]

@LindseyB the same as Azure OpenAI o1-preview, right?

Prototyping with AI models

Rate limits

Rate limit tier	Rate limits	Copilot Free	Copilot Pro	Copilot Business	Copilot Enterprise
...	...	...	...	...	...
Azure OpenAI o1-preview	Requests per minute	Not applicable	1	2	2
Azure OpenAI o1-preview	Requests per day	Not applicable	👉 8 👈	10	12
Azure OpenAI o1-preview	Tokens per request	Not applicable	4000 in, 4000 out	4000 in, 4000 out	4000 in, 8000 out
Azure OpenAI o1-preview	Concurrent requests	Not applicable	1	1	1
...	...	...	...	...	...

---

Will this table be updated to include Azure DeepSeek R1 as new rows, considering it appears as Custom on the Rate Limit Tier in the GitHub models playground?

[HillPhelmuth]

I'm also getting a rate limit error that doesn't match the high rate limit numbers for Github Pro. Based on the error message, it seems to be applying the o1-preview rate limits.

[429 Too Many Requests]: {"error":{"code":"RateLimitReached","message":"Rate limit of 8 per 86400s exceeded for UserByModelByDay. Please wait 83165 seconds before retrying.","details":"Rate limit of 8 per 86400s exceeded for UserByModelByDay. Please wait 83165 seconds before retrying."}} 

[LindseyB]

Updating for a correction: the documentation for the model is incorrect DeepSeek is currently free tier not high and has the free rate limit which is what you are seeing. We will be making an adjustment to our documentation sorry about that.

[llightcb]

thanks for your playground. i can play a lot within it. one single question/prompt/preview. this is a gift for all of us. this is really helpful. thank you so much. and it's blazing fast too. holy moly.

[Lyonsclay]

When I try the code offered in the 'Use this model' tab; https://github.com/marketplace/models/azureml-deepseek/DeepSeek-R1/playground. I get an error;

args -- ('What is the capital of France?',)
Traceback (most recent call last):
  File "/Users/claymorton/developer/github-models/deepseek.py", line 20, in <module>
    UserMessage("What is the capital of France?"),
    ~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/claymorton/developer/venv/lib/python3.13/site-packages/azure/ai/inference/models/_models.py", line 1291, in __init__
    super().__init__(*args, role=ChatRole.USER, **kwargs)
    ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/claymorton/developer/venv/lib/python3.13/site-packages/azure/ai/inference/models/_models.py", line 55, in __init__
    super().__init__(*args, **kwargs)
    ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
  File "/Users/claymorton/developer/venv/lib/python3.13/site-packages/azure/ai/inference/_model_base.py", line 495, in __init__
    {k: _create_value(_get_rest_field(self._attr_to_rest_field, k), v) for k, v in args[0].items()}
                                                                                   ^^^^^^^^^^^^^
AttributeError: 'str' object has no attribute 'items'

It appears the args aren't getting passed through correctly. The same code works fine for the Llama-3.3-70B-Instruct and Llama-3.3-70B-Instruct models.