GPT-4 (Mar 2023) (OpenAI) and Nemotron-4 340B (NVIDIA) are both frontier AI models. GPT-4 (Mar 2023) was published in March 2023 and Nemotron-4 340B in June 2024.
GPT-4 (Mar 2023) was trained on 2.1 x 10^25 FLOP, about 1.2x the compute of Nemotron-4 340B at 1.8 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 (Mar 2023) | Nemotron-4 340B |
|---|---|---|
| Organization | OpenAI | NVIDIA |
| Published | Mar 15, 2023 | Jun 14, 2024 |
| Training compute | 2.1 x 10^25 FLOP | 1.8 x 10^25 FLOP |
| Parameters | 1.8T | 340B |
| Dataset size | 5.4T | 9T |
| Training hardware | NVIDIA A100 SXM4 40 GB | NVIDIA H100 SXM5 80GB |
| Chips used | 25,000 | 6,144 |
| Training time | 2.3K h | 2.2K h |
| Training cost (2023 USD) | $37M | $21M |
| Training power draw | 19.9 MW | 8.5 MW |
| Accessibility | API access | Open weights (unrestricted) |
| Open weights | No | Yes |
| Country | United States | United States |
