Head to head
Grok 4 vs Llama 4 Behemoth (preview)
5×10²⁶Training compute (FLOP)5.2×10²⁵
$388MTraining cost$45M
Grok 4 (xAI) and Llama 4 Behemoth (preview) (Meta AI) are both frontier AI models. Grok 4 was published in July 2025 and Llama 4 Behemoth (preview) in April 2025.
Grok 4 was trained on 5×10²⁶ FLOP, about 9.6x the compute of Llama 4 Behemoth (preview) at 5.2×10²⁵ 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
Meta AI
Jul 9, 2025
Published
Apr 5, 2025
5×10²⁶ FLOP
Training compute9.6x
5.2×10²⁵ FLOP
3T
Parameters1.5x
2T
--
Dataset size
30T
--
Training hardware
NVIDIA H100 SXM5 80GB
200,000
Chips used6.3x
32,000
$388M
Training cost (2023 USD)8.7x
$45M
--
Training power draw
43.9 MW
API access
Accessibility
Unreleased
No
Open weights
No
United States
Country
United States
Related comparisons
Grok 4 vs GPT-4.5Grok 4 vs Gemini 1.0 UltraGrok 4 vs Llama Nemotron Ultra 253BGrok 4 vs Llama 3.1-405BGrok 4 vs Claude 3.5 SonnetGrok 4 vs Nemotron-4 340BGrok 4 vs Inflection-2Grok 4 vs Amazon TitanGrok 4 vs Claude 2Grok 4 vs Falcon-180BGPT-4.5 vs Llama 4 Behemoth (preview)Grok 3 vs Llama 4 Behemoth (preview)
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.