Nemotron 3 Super Equivalent Model Continues Training
scaling01 · x · 2026-07-12
The post hints at "more to come," pointing to a "Nemotron 3 Super equivalent model." Background info reveals the team believes that mastering the end-to-end data pipeline, data sources, and infrastructure requires more than just fine-tuning someone else's checkpoint.
They spent about 5 months preparing for this project, with around 20 engineers involved before formal pretraining began. They plan to train an improved Nemotron 3 Super equivalent model with a better data mixing strategy and will share more details soon.
More from Models
- Gary Marcus says LLM math skills are like knowing only a car’s engine size — GaryMarcus · 2026-07-22
- OpenAI’s Codex + GPT-5.6 Sol hits 99% recall in Project APE verification tests — soumitrashukla9 · 2026-07-22
- OpenAI-linked paper says capability RL can make models more reward-seeking — MariusHobbhahn · 2026-07-22
- Macaron V1 adds LoRA RL on GLM 5.2 and claims SOTA benchmark gains — Xianbao_QIAN · 2026-07-22
- OpenAI rolls out voice in GPT-Live, but the UI obscures search and reasoning — Graham_dePenros · 2026-07-22
- Moonshot’s Kimi K3 sets a new open-weights ECI record at 156 — scaling01 · 2026-07-22