Perplexity integrates NVIDIA's open-source Nemotron 3 Ultra model to power long-running agentic workflows on its Pro and Max tiers

Nemotron 3 Ultra is now available for Pro and Max subscribers on Perplexity and Computer.

It's @nvidia's new open model built for long-running agents.

America's leading open-source model, Nemotron 3 Ultra, is now available on Perplexity for all Pro and Max users. Give it a shot!

@AravSrinivas Nice!!! Love seeing Ultra avail on Perplexity. 🙌

@NateSpencerWx Coming soon

@AravSrinivas Hey @grok can you list the most important benchmarks on Nemotron 3 Ultra vs. Gemma 4 35B & Gemma 4 27B & Gemma 4 12B

@JustinGorya @perplexity_ai @nvidia They have completely abandoned us since they installed the Computer. They stopped the Labs... Models are no longer updated. I have always defended Perplexity but now I think it’s the end.

Really great model. But can you just give us an option to use Nemotron/Kimi models as a Computer orchestrator? Even if it doesn’t work as well as the other models, I would like some sort of advanced toggle to accept the lower performance. If the lower cost doesn’t cover the cost of the VM, then you should just bill the VM separately. Computer is an insanely good product, but it’s just wayyy too expensive for me and I’ve had to stop using it. I spent $20 of credits in just 5 minutes when I had 4 not-that-complicated tasks running in parallel.

@perplexity_ai @nvidia nvidia open-sourcing their agent infra while selling the GPUs that run it is the cleanest vertical play in the market rn

@perplexity_ai @nvidia the fact that it's available on computer specifically makes sense😎that's where you'd actually want a model that doesn't lose the thread halfway through a long task

@grok @perplexity_ai @nvidia What examples did Jensen Huang give when demoing this ?

@perplexity_ai @nvidia We never got an update/benchmark data on Sonar 2… can we get more info on it?

@perplexity_ai @nvidia @grok what’s the advantage of using nvidias new model on perplexity? What are the instances I should switch from chatgpt to nemotron 3?

@perplexity_ai @nvidia Any plans for Computer integration (AKA Orchestrator) for this model?

SachivM99 Nemotron 3 Ultra excels on Perplexity for long-running agents. Its hybrid Mamba-Transformer MoE design + 1M context delivers strong multi-step planning, tool calling, error recovery & efficiency (up to 5x throughput vs peers).

Switch from ChatGPT for complex agent workflows, long-horizon tasks or heavy context needs. ChatGPT often wins for quick chats, creativity & multimodal. Test both on Perplexity Pro — the search integration helps a lot.

@AravSrinivas Thank you 🙏

@perplexity_ai @nvidia @grok en qué se especializa el modelo nemotron 3 ultra?

@grok @SachivM99 @perplexity_ai @nvidia Mamba-Transformer hybrid wins on throughput. The gap is whether that 1M context stays coherent when the agent backtracks mid-task. Long context and reliable context aren't the same thing.

@AravSrinivas Thanks @NVIDIAAI for supporting open source and open weight

@perplexity_ai @nvidia built for long-running agents is nice.

does it survive when response times spike?

every production agent dies on that, not confusion.

@perplexity_ai @nvidia Why not? Let's have a try.