AI Potluck
Model components / Fine-tuned / chat models

Nemotron 3

NVIDIA

NVIDIA's first-party open-weight instruction-tuned model family for agentic reasoning, post-trained with SFT + RL on top of Nemotron Base. Current SKUs are Nemotron 3 Nano (30B total / 3B active hybrid MoE, March–April 2026) and Nemotron 3 Super (120B / 12B active Mamba-Transformer MoE, March 11, 2026); Nemotron 3 Ultra (550B total / 55B active) shipped 3 June 2026. Differentiates on (a) open training data and recipes published alongside the weights, (b) 1M-token context with strong RULER scores (Super 91.8 @ 1M), and (c) being NVIDIA's reference post-trained line for NeMo and NIM, optimized end-to-end against NVIDIA hardware. Adoption signal (Hub, 2026-07-29): ~994K monthly downloads on Nano BF16, 855K on Super BF16 and 339K on Ultra BF16, plus free hosting on OpenRouter and build.nvidia.com.

The Nemotron 3 family: Nano (30B total / 3B active hybrid MoE), Super (120B / 12B active hybrid Mamba-Transformer MoE, 1M context, 11 Mar 2026 at GTC) and Ultra (550B / 55B active, 3 Jun 2026). Designed for agentic/conversational use. Held a duplicate record in base_pretrained until 2026-07-29; the two were merged here because the family distributes post-trained weights only. The sole -Base repo on the Hub is nvidia/nemotron-3-8b-base-4k, created 2023-11-14 with a single download, which is the earlier 8B generation and not this family. Per the #114 convention, the observable artifact decides the category. BF16 originals are declared above; the NVFP4/FP8/GGUF quantizations are the same weights repackaged. An earlier note recorded the instruct repo as gated/401 -- all three resolve ungated as of 2026-07-29.

Openness

4 high confidence
4.0
weights
open
data
open(~40M post-training/SFT+RL samples released as Nemotron-SFT-Data/Nemotron-RL-Data
code
open(training recipes + RL env configs)
license
NVIDIA-Nemotron-Open-Model-License(permissive,commercial,non-OSI)

Unusually open for post-trained models: NVIDIA releases weights, the post-training data and the recipes, but under the non-OSI NVIDIA Open Model License, so open_weights not open_source. This category scores the post-training data, which is genuinely released. The pretraining side is thinner than NVIDIA's "fully open" framing suggests -- Nemotron-Pretraining-Specialized-v1.2 ships 53.6GB of synthetic data, but Nemotron-Pretraining-Code-v3 ships metadata parquet describing the code corpus rather than the corpus, so the earlier "~10T-token pretraining released" wording overstated it and has been narrowed. Score is unaffected: the post-training answer is what this category asks for.

Adoption

4 high confidence
4.0

Corrected from 3/reported_traction to 4/usage_volume on 2026-07-29, and this is the first score on the map re-banded off a machine signal rather than a hand-read figure. The record previously said no hard download count was available because the instruct repo was gated; all three shipped repos resolve ungated at HTTP 200. Summing the family's distributed SKUs, which is what the category's adoption recipe specifies: Nano BF16 994,374 + Super BF16 855,265 + Ultra BF16 339,360 = 2,188,999 monthly downloads, squarely in the 1M-10M band. The NVFP4/FP8/GGUF quantizations are excluded to avoid double counting the same weights. Also distributed free via OpenRouter (nemotron-3-super-120b-a12b:free) and NVIDIA build/NIM, which the band does not capture, so 4 is a floor.

Capability

5 high confidence
5.0

Frontier-tier among open instruct models on the Super SKU. SWE-bench Verified 60.47 confirmed as highest open-weight score via NVIDIA technical report + developer blog. Nano is much smaller and would score lower, but the family's headline SKU defines the score.

Unchanged since 2026-07-29 (last edited, not re-checked)