Llama 3.1-405B (Meta AI) and U-PaLM (540B) (Google) are both frontier AI models. Llama 3.1-405B was published in July 2024 and U-PaLM (540B) in October 2022.
Llama 3.1-405B was trained on 3.8 x 10^25 FLOP, about 15.0x the compute of U-PaLM (540B) at 2.5 x 10^24 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 | Llama 3.1-405B | U-PaLM (540B) |
|---|---|---|
| Organization | Meta AI | |
| Published | Jul 23, 2024 | Oct 20, 2022 |
| Training compute | 3.8 x 10^25 FLOP | 2.5 x 10^24 FLOP |
| Parameters | 405B | 540B |
| Dataset size | 15.6T | 1.3B |
| Training hardware | NVIDIA H100 SXM5 80GB | Google TPU v4 |
| Chips used | 16,384 | 512 |
| Training time | 2.1K h | 120 h |
| Training cost (2023 USD) | $53M | — |
| Training power draw | 22.6 MW | 348.3 kW |
| Accessibility | Open weights (restricted use) | Unreleased |
| Open weights | Yes | No |
| Country | United States | United States |
