DeepSeek R1 has been released under the MIT license!
Hoping that SambaNova considers adding this model.
7 Likes
Hi,
Thank you for your suggestion regarding DeepSeek R1. We truly appreciate you bringing this up.
We’ll pass this along to our team for further discussion and evaluation. It’s definitely worth exploring, and we’ll take a closer look at the feasibility and how it might align with our offerings.
Thanks again for your valuable feedback, and we’ll keep you updated if there are any developments!
Regards,
Prajwal
2 Likes
Us llm hardware or api providers doesnt want to use chinese models.
Same with the groq and cerebras.
They dont want the us devs to build on top of chinese models.
1 Like
@vibright10 @NB2025 and @kollikiran456
SambaNova Systems recognizes the benefit and achievements of Open Source models globally. Bringing new models to the platform is a continuous product management and engineering process but we cannot commit to specific model deliveries nor release dates . This being said your voices are heard and your desires have been conveyed to product management. As we know more we will let the community know.
-Coby
That is for our SambaNova Suite product . I can put you in contact with enterprise sales if your organization is looking for an on prem or hosted , private, training/inference solution .
-Coby
1 Like
I’m just a developer looking to implement it in my app, still not released
Time to run it on a local system: Reddit - Dive into anything
Unsloth released deepseek r1 1.5 bit.
Can be run on “cpu” @ 3 tokens/sec.