Flan-PaLM 540B (Google) and Inflection-2 (Inflection AI) are both frontier AI models. Flan-PaLM 540B was published in October 2022 and Inflection-2 in November 2023.
Inflection-2 was trained on 10^25 FLOP, about 3.9x 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 | Flan-PaLM 540B | Inflection-2 |
|---|---|---|
| Organization | Inflection AI | |
| Published | Oct 20, 2022 | Nov 22, 2023 |
| Training compute | 2.5 x 10^24 FLOP | 10^25 FLOP |
| Parameters | 540B | — |
| Dataset size | 1.4B | — |
| Training hardware | Google TPU v4 | NVIDIA H100 SXM5 80GB |
| Chips used | 512 | 5,000 |
| Training time | 37 h | — |
| Training cost (2023 USD) | — | $13M |
| Training power draw | 348.3 kW | 6.9 MW |
| Accessibility | Unreleased | Hosted access (no API) |
| Open weights | No | No |
| Country | United States | United States |
