GPT-4 (Mar 2023) (OpenAI) and Grok 3 (xAI) are both frontier AI models. GPT-4 (Mar 2023) was published in March 2023 and Grok 3 in February 2025.
Grok 3 was trained on 3.5 x 10^26 FLOP, about 16.7x the compute of GPT-4 (Mar 2023) at 2.1 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 | GPT-4 (Mar 2023) | Grok 3 |
|---|---|---|
| Organization | OpenAI | xAI |
| Published | Mar 15, 2023 | Feb 17, 2025 |
| Training compute | 2.1 x 10^25 FLOP | 3.5 x 10^26 FLOP |
| Parameters | 1.8T | 3T |
| Dataset size | 5.4T | — |
| Training hardware | NVIDIA A100 SXM4 40 GB | NVIDIA H100 SXM5 80GB |
| Chips used | 25,000 | 80,000 |
| Training time | 2.3K h | 2.2K h |
| Training cost (2023 USD) | $37M | $218M |
| Training power draw | 19.9 MW | 109.9 MW |
| Accessibility | API access | API access |
| Open weights | No | No |
| Country | United States | United States |
| Benchmark | GPT-4 (Mar 2023) | Grok 3 |
|---|---|---|
| Epoch Capabilities Index(ECI Score) | 125.9 | 139 |
