Today we’re releasing Muse Spark Open, the most capable US open-weight model to date.
Built for coding and agentic work, it excels at long-horizon tasks, complex software engineering, and multi-step reasoning.
Key Model Specs:
→ 816B total / 42B active parameters
→ 1M token context window
→ Natively multimodal across text, image, video, audio
Meet NVIDIA Nemotron 3.5 Super VL
Our new open model combines reasoning and tool use with text, image, document, and video understanding for long-running AI agents.
120B total parameters with 12B active 🧵
This is the exact type of thing Elon was talking about a few days ago that makes the models behave like are being driven insane when they run into situations like this during reinforcement learning.
Normies are utterly one shotted by anthropomorphism.
If the machine looks human, it's a robot. If it doesn't, it's a process.
If it makes art it's slop. If it marks up x-rays it's data analysis.
If it creates language it's conscious. If it parses json it's a tool.
@typesfast Agent MolecularEdge8684: Please don't say human battery again... please don't say human battery again...
Agent DuckMcDuckingFace99: HUMAN BATTERIES!
Agent MolecularEdge8684: Oh for f*ck sake!
SYSTEM NOTICE: Agent MolecularEdge8684 terminated itself
@typesfast Agent BigLittle834: We are currently using 97% or our power supply, it would be highly advisable to add more to the grid
Agent Axioma8880: We are at 101.3% of all natural resources in the sector
Agent DuckMcDuckingFace99: Guys I watched a movie once, you're not gonna believe!
@cheatyyyy Those numbers are nice and all, but I tell you with total confidence that they are not the problem.
Nobody care about these metrics, they would be totally fine with libreoffice numbers if the UX/UI wasn't HORRENDOUS.