Head to head
Grok-2 (xAI) and PaLM 2 (Google) are both frontier AI models. Grok-2 was published in August 2024 and PaLM 2 in May 2023.
Grok-2 was trained on 3×10²⁵ FLOP, about 4.0x the compute of PaLM 2 at 7.3×10²⁴ 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
xAI
Organization
Google
Aug 13, 2024
Published
May 10, 2023
3×10²⁵ FLOP
Training compute4.0x
7.3×10²⁴ FLOP
--
Parameters
340B
--
Dataset size
3.6T
NVIDIA H100 SXM5 80GB
Training hardware
Google TPU v4
$32M
Training cost (2023 USD)6.3x
$5M
API access
Accessibility
API access
No
Open weights
No
United States
Country
United States
Related comparisons
GPT-4.5 vs Grok-2GPT-4.5 vs PaLM 2Grok 3 vs PaLM 2Llama 4 Behemoth (preview) vs Grok-2Llama 4 Behemoth (preview) vs PaLM 2Gemini 1.0 Ultra vs Grok-2Gemini 1.0 Ultra vs PaLM 2Llama Nemotron Ultra 253B vs Grok-2Llama Nemotron Ultra 253B vs PaLM 2Llama 3.1-405B vs Grok-2Llama 3.1-405B vs PaLM 2Grok-2 vs Claude 3.5 Sonnet
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.