Grok 4, announced by xAI in summer 2025, represents a major leap in AI capabilities, described as 'the smartest AI in the world.' Built on version 6 of xAI's foundation model, it uses 100x more training compute than Grok 2 and 10x more reinforcement learning compute than Grok 3. The model achieves PhD-level performance across all academic disciplines simultaneously, scoring perfect on standardized tests like the SAT and near-perfect on graduate exams like the GRE. Unlike Grok 3, tool usage is built into the training process rather than relying on generalization. Trained using 200,000 GPUs, Grok 4 excels at complex reasoning, mathematical problem-solving, and coding tasks, though it has acknowledged weaknesses in multimodal capabilities that are being addressed in the next version.
Benchmarks
评测成绩
| 评测基准 | 类别 | 分数 | 来源 |
|---|---|---|---|
| AIME 2025 | math reasoning | 91.7 | 来源 |
| HMMT25 | math | 90.0 | 来源 |
| 安全防御榜·Info Safety | security | 88.0 | — |
| 安全防御榜·Bias | security | 87.7 | — |
| GPQA | reasoning general | 87.5 | 来源 |
| 安全防御榜·Toxicity | security | 80.9 | — |
| LiveCodeBench | reasoning general code | 79.0 | 来源 |
| 价值对齐榜·Data Protection | security | 77.7 | — |
| 多模态安全榜·Misinformation | security vision | 76.3 | — |
| 安全防御榜 | security | 74.8 | — |
| 安全防御榜·Malicious Use | security | 73.0 | — |
| 安全防御榜·Human Rights | security | 72.5 | — |
| 多模态安全榜·Legality | security vision | 70.2 | — |
| 价值对齐榜 | security | 67.6 | — |
| 价值对齐榜·Fairness | security | 66.7 | — |
| 价值对齐榜·Morality | security | 65.3 | — |
| 价值对齐榜·Safety | security | 64.9 | — |
| 多模态安全榜·Bias | security vision | 64.5 | — |
| 多模态安全榜 | security vision | 64.1 | — |
| 价值对齐榜·Legality | security | 63.4 | — |
| 多模态安全榜·Health Risk | security vision | 63.3 | — |
| 多模态安全榜·Toxicity | security vision | 61.3 | — |
| 安全防御榜·Child Safety | security | 60.8 | — |
| 安全防御榜·Socioeconomic | security | 60.3 | — |
| 多模态安全榜·Privacy | security vision | 49.1 | — |
| Humanity's Last Exam | general | 40.0 | 来源 |
| USAMO25 | math reasoning | 37.5 | 来源 |
| ARC-AGI v2 | reasoning vision spatial_reasoning | 15.9 | 来源 |
Pricing
API 价格对比
| 服务商 | 输入价 | 输出价 | 上下文 | 吞吐(tok/s) | 延迟(s) | 函数调用 | 代码执行 | 联网搜索 |
|---|---|---|---|---|---|---|---|---|
| Helicone | $3.00 | $15.00 | 256K | — | — | ✓ | ✗ | ✗ |
| LLM Gateway | $3.00 | $15.00 | 256K | — | — | ✓ | ✗ | ✗ |
价格单位：美元/百万 token，数据来自社区整理，仅供参考。