Grok 3 (xAI) and Llama 3.1-405B (Meta AI) are both frontier AI models. Grok 3 was published in February 2025 and Llama 3.1-405B in July 2024.
Grok 3 was trained on 3.5 x 10^26 FLOP, about 9.2x the compute of Llama 3.1-405B at 3.8 x 10^25 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.
Grok 3 wins the single benchmark both were scored on. Benchmark counts are a crude scoreboard — the evaluations differ wildly in what they measure and in how saturated they are — so the per-benchmark table below matters more than the tally.
| Field | Grok 3 | Llama 3.1-405B |
|---|---|---|
| Organization | xAI | Meta AI |
| Published | Feb 17, 2025 | Jul 23, 2024 |
| Training compute | 3.5 x 10^26 FLOP | 3.8 x 10^25 FLOP |
| Parameters | 3T | 405B |
| Dataset size | — | 15.6T |
| Training hardware | NVIDIA H100 SXM5 80GB | NVIDIA H100 SXM5 80GB |
| Chips used | 80,000 | 16,384 |
| Training time | 2.2K h | 2.1K h |
| Training cost (2023 USD) | $218M | $53M |
| Training power draw | 109.9 MW | 22.6 MW |
| Accessibility | API access | Open weights (restricted use) |
| Open weights | No | Yes |
| Country | United States | United States |
| Benchmark | Grok 3 | Llama 3.1-405B |
|---|---|---|
| Epoch Capabilities Index(ECI Score) | 139 | 129 |
