Nvidia Launches Family of Open Reasoning AI Models: OpenReasoning Nemotron | Hacker News Reader

Nvidia Launches Family of Open Reasoning AI Models: OpenReasoning Nemotron

nvidianews.nvidia.com

84 pointskristianp18 comments

Is this another repurposing and bastardization of “Open” or are these actually open? Should I even be asking?

For any actual openness in released models the dataset they're trained on would have to be released as well, so yeah, it is a bastardisation.

It is in nvidia's interest to commoditize one's complement. These models make owning and using nvidia hardware more attractive - being open and all.

There's no incentive to "hide" the sauce.

It's under a Creative Commons Attribution license (CC-BY). That's about as open as it gets.

NVIDIA unveiled OpenReasoning-Nemotron, a quartet of distilled reasoning models with 1.5B, 7B, 14B, and 32B parameters, all derived from the 671B-parameter DeepSeek R1 0528. [1]

[1] https://www.techpowerup.com/339089/nvidia-brings-reasoning-m...

This is from March. Why is it being re-posted?

Gotta pump NVDA shares higher

That nvidia announcement was from March, but I think something has been released recently, as there were a bunch of news stories about a week ago. This blog post [1] was released on July 18th, for example.

[1] https://huggingface.co/blog/nvidia/openreasoning-nemotron

Crazy how a few months make such a difference.

AI is in 5x historical-tech-speed-mode. So, this is kind of like posting about an iPhone 15.

lol! Claude told me last night, to a question about MCP confusion, that I was experiencing “AI dog years”!

that's funny cause a lot of AI's dont' even know yet what MCP is even

I remember “back in the day” when there were under 100 servers listed on glama/gh and thinking “wow this is growing quick!”. lol. ¯\_(ツ)_/¯

the protocol spec itself has evolved rapidly as anthropic seems to have gone from side-project to taking it way more seriously as it took off

most of the servers are meh, but some of the stuff is cool, security remains an issue but can't ask a protocol to solve everything i guess

lol!! but also driving this is the fact is that each new thing speeds up the development of the next thing so 5x is more like (e^t)x

https://huggingface.co/collections/nvidia/openreasoning-nemo...

The model card, and the "you have to be authed" image pull instructions: https://build.nvidia.com/nvidia/llama-3_1-nemotron-70b-instr...

March 18, 2025?

Reply on news.ycombinator.com