Traictory tracks 11 models from NVIDIA. The earliest entry is Llama 3.1 Nemotron 70B Instruct from October 2024; the most recent is Nemotron Labs Dory 1.0 (4B), released in October 2026. 7 of them shipped in the last twelve months.
The strongest NVIDIA model in our data is Llama-3.3 Nemotron Super 49B v1, with an average benchmark score of 81.0% across 7 published results, including 66.7% on GPQA and 96.6% on MATH-500. Across the 11 NVIDIA models with benchmark data the average is 61.7%.
3 of the models are available through an API, with input pricing from $0.05 to $0.20 per million tokens; Nemotron 3 Nano (30B A3B) is the cheapest. The largest context window is 262K tokens on Nemotron 3.5 Super VL (120B A12B). 2 of the 11 models accept images or other non-text input. 9 models ship with downloadable weights; the rest are API-only.
NVIDIA Models
Other Companies
NVIDIA model catalog updated in 2026. See model comparisons and best models for coding.