FLAN-UL2
Similar to FLAN-T5, FLAN-UL2 is a model based on Google's popular T5 architecture with an upgraded pre-training procedure dubbed UL2. On most NLU benchmarks, FLAN-UL2 outperforms FLAN-T5 by a significant margin.
A side-by-side comparison of release history, capabilities, licensing and model sizes.
| Attribute |
 FLAN-UL2
|
Nemotron |
|---|---|---|
| Developer | NVIDIA | |
| Current generation | – | Nemotron 3 Ultra |
| Initial release | 2023-03-03 | 2025-12-15 |
| Instruct tuned | Yes | Yes |
| Coding | No | Yes |
| Finetuning | Yes | Yes |
| Open source | Yes | Yes |
| License | Apache 2.0 | OpenMDW-1.1 |
| Model sizes | 20B | 550B (55B active) |
| Links |
Similar to FLAN-T5, FLAN-UL2 is a model based on Google's popular T5 architecture with an upgraded pre-training procedure dubbed UL2. On most NLU benchmarks, FLAN-UL2 outperforms FLAN-T5 by a significant margin.
Nemotron is NVIDIA's open model family for efficient reasoning, coding, long-context analysis, and agent orchestration. Nemotron 3 Ultra is the current flagship.
NVIDIA introduced the Nemotron 3 family in December 2025 with Nano, Super, and Ultra tiers. Nemotron 3 Ultra, released in June 2026, is a 550B-parameter hybrid Mamba-Transformer mixture-of-experts model with 55B active parameters and a one-million-token context window. It is optimized for long-running agent workflows, tool use, code, multilingual reasoning, and high-stakes retrieval. NVIDIA publishes the weights, training data, recipes, and fine-tuning workflows under the OpenMDW-1.1 license.