samsja

samsja

833 Twitter followers

Follow

Research Engineer at @PrimeIntellect, previously training llm at @NyonicAI, maintainer of @docarray

Posts

can't believe we are living in the timeline where terminal, tui, markdown and git are going to take over the world

we just merged slurm support into prime-rl, a single command that render and start a slurm script to start a large moe training over 512+ gpus we also have k8s support

Today we’re releasing Trinity Large, a 400B MoE LLM with 13B active parameters, trained over 17T tokens The base model is on par with GLM-4.5 Base, while being significantly faster at inference because it’s sparser and hybrid The architecture we picked is one of my favorites:

Prime Intellect

@PrimeIntellect

We're excited to introduce @arcee_ai's Trinity Large model. An open 400B parameter Mixture of Experts model, delivering frontier-level performance with only 13B active parameters. Trained in collaboration between Arcee, Datology and Prime Intellect. x.com/arcee_ai/statu…

Decentralized training is a great solution for Neo Cloud to repurpose their old generation of hardware (google has been doing this for a while with their TPU)

2️⃣ Blackwell Mass Deployment has Started Mass production and initial deployments of @nvidia Blackwell (B100/GB200) are ramping fast. These next-gen GPUs offer better performance per watt and lower TCO, making H100s yesterday’s news for neocloud customers. As a result, many

Loading..