Falcon-180B (Technology Innovation Institute) and Flan-PaLM 540B (Google) are both frontier AI models. Falcon-180B was published in September 2023 and Flan-PaLM 540B in October 2022.
Falcon-180B was trained on 3.8 x 10^24 FLOP, about 1.5x the compute of Flan-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 | Falcon-180B | Flan-PaLM 540B |
|---|---|---|
| Organization | Technology Innovation Institute | |
| Published | Sep 6, 2023 | Oct 20, 2022 |
| Training compute | 3.8 x 10^24 FLOP | 2.5 x 10^24 FLOP |
| Parameters | 180B | 540B |
| Dataset size | 3.5T | 1.4B |
| Training hardware | NVIDIA A100 SXM4 40 GB | Google TPU v4 |
| Chips used | 4,096 | 512 |
| Training time | 4.3K h | 37 h |
| Training cost (2023 USD) | $11M | — |
| Training power draw | 3.3 MW | 348.3 kW |
| Accessibility | Open weights (restricted use) | Unreleased |
| Open weights | Yes | No |
| Country | United Arab Emirates | United States |
