StepFun's flagship agentic model, a 600B-parameter MoE with 27B active.
NVIDIA's open frontier reasoning model for orchestration, coding, and deep research.
No shared demos for these models yet.
No captured outputs for these models yet.
Both Step 5 Preview and Nemotron 3 Ultra are capable AI models. The best choice depends on your specific use case: consider pricing, context window, speed, and which capabilities (vision, reasoning, audio) you need.
Nemotron 3 Ultra has a lower blended cost. Step 5 Preview: $1.00 input / $2.70 output. Nemotron 3 Ultra: $0.60 input / $2.40 output.
Step 5 Preview has a larger context window: Step 5 Preview supports 1M tokens vs Nemotron 3 Ultra at 203K tokens.
Step 5 Preview supports vision/image input, but Nemotron 3 Ultra does not.
Key differences: only Step 5 Preview supports vision input. Compare full specs on this page.
If cost is your priority, choose the cheaper option. If you need the highest intelligence for complex tasks, pick the higher-scoring model. For long documents or codebases, choose the larger context window. You can try both Step 5 Preview and Nemotron 3 Ultra for free on idapt.app to see which performs better for your specific needs.