Nemotron 4 is getting close 🟩 • NVIDIA's TensorRT-LLM repo shows an agent onboarding "nemotron4-nano-preview" end to end on GB300 • FlashInfer merged kernels for "Nemotron 4 Ultra" and "Nemotron 4 Nano" (Gated DeltaProduct + multi-token prediction), then scrubbed the name
- Status
- Leaked, not yet released
- Maker
- NVIDIA
- Found in
- github.com github.com
- Detected
- Posted to X
- 0 s after detection
- Model ID
Nemotron 4 (Nano / Ultra)