GPT-4 (Jun 2023) (OpenAI) and Llama Nemotron Ultra 253B (NVIDIA) are both frontier AI models. GPT-4 (Jun 2023) was published in June 2023 and Llama Nemotron Ultra 253B in March 2025.
Llama Nemotron Ultra 253B was trained on 3.9 x 10^25 FLOP, about 1.9x the compute of GPT-4 (Jun 2023) at 2.1 x 10^25 FLOP. Training compute is the closest available proxy for how much was invested in a model, though it says nothing on its own about how well that compute was spent.
These two models share no benchmark on which both have been scored, so no direct performance comparison is possible here. The specification table below is a comparison of inputs, not of results.
| Field | GPT-4 (Jun 2023) | Llama Nemotron Ultra 253B |
|---|---|---|
| Organization | OpenAI | NVIDIA |
| Published | Jun 13, 2023 | Mar 18, 2025 |
| Training compute | 2.1 x 10^25 FLOP | 3.9 x 10^25 FLOP |
| Parameters | 1.8T | 253B |
| Dataset size | 5.4T | 603B |
| Training hardware | NVIDIA A100 SXM4 40 GB | — |
| Chips used | 25,000 | — |
| Training time | 2.3K h | — |
| Training power draw | 19.9 MW | — |
| Accessibility | API access | Open weights (restricted use) |
| Open weights | No | Yes |
| Country | United States | United States |
