Head to head
Inflection-2 vs PaLM (540B)
10²⁵Training compute (FLOP)2.5×10²⁴
$13MTraining cost$3M
Inflection-2 (Inflection AI) and PaLM (540B) (Google Research) are both frontier AI models. Inflection-2 was published in November 2023 and PaLM (540B) in April 2022.
Inflection-2 was trained on 10²⁵ FLOP, about 4.0x the compute of 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
Inflection AI
Organization
Google Research
Nov 22, 2023
Published
Apr 4, 2022
10²⁵ FLOP
Training compute4.0x
2.5×10²⁴ FLOP
--
Parameters
540.4B
--
Dataset size
780B
NVIDIA H100 SXM5 80GB
Training hardware
Google TPU v4
5,000
Chips used1.2x
6,144
--
Training time
1,536 h
$13M
Training cost (2023 USD)4.4x
$3M
6.9 MW
Training power draw
4.2 MW
Hosted access (no API)
Accessibility
Unreleased
No
Open weights
No
United States
Country
United States
Related comparisons
Grok 4 vs Inflection-2GPT-4.5 vs Inflection-2Grok 3 vs Inflection-2Llama 4 Behemoth (preview) vs Inflection-2Gemini 1.0 Ultra vs Inflection-2Gemini 1.0 Ultra vs PaLM (540B)Llama Nemotron Ultra 253B vs Inflection-2Llama 3.1-405B vs Inflection-2Grok-2 vs Inflection-2Claude 3.5 Sonnet vs Inflection-2GPT-4 (Jun 2023) vs Inflection-2GPT-4 (Jun 2023) vs PaLM (540B)
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.