实时
Head to head

Flan-PaLM 540B vs GPT-4 (Mar 2023)

Google
Flan-PaLM 540B
October 2022
vs
OpenAI
GPT-4 (Mar 2023)
March 2023
2.5×10²⁴Training compute (FLOP)2.1×10²⁵
--Training cost$37M

Flan-PaLM 540B (Google) and GPT-4 (Mar 2023) (OpenAI) are both frontier AI models. Flan-PaLM 540B was published in October 2022 and GPT-4 (Mar 2023) in March 2023.

GPT-4 (Mar 2023) was trained on 2.1×10²⁵ FLOP, about 8.3x the compute of Flan-PaLM 540B at 2.5×10²⁴ 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.

Specifications
Google
Organization
OpenAI
Oct 20, 2022
Published
Mar 15, 2023
2.5×10²⁴ FLOP
Training compute8.3x
2.1×10²⁵ FLOP
540B
Parameters3.3x
1.8T
1.4B
Dataset size3869x
5.4T
Google TPU v4
Training hardware
NVIDIA A100 SXM4 40 GB
512
Chips used49x
25,000
37 h
Training time
2,280 h
--
Training cost (2023 USD)
$37M
348.3 kW
Training power draw
19.9 MW
Unreleased
Accessibility
API access
No
Open weights
No
United States
Country
United States
Related comparisons
SourceEpoch AI, 'AI Models'. Published online at epoch.ai. Retrieved 2026-07-29 from https://epoch.ai/data/ai-models. Licensed under CC BY 4.0.