Grok 4.7 is SpaceXAI's flagship model for coding, agentic tasks, and knowledge work, succeeding Grok 4.6. It is particularly strong at long-running software engineering tasks, verifying its own work, and managing long context, and it improves on its predecessor at professional knowledge work such as drafting documents and presentations.
The model was trained with a longer reinforcement learning run weighted toward problems that take many hours to complete, and natively understands the Grok Bot harness for conversational tasks. It ships with a new safeguard stack that pairs strong jailbreak resistance with low refusal rates for legitimate cybersecurity and biology work. SpaceXAI's reported benchmark results use the xhigh reasoning effort.
提供商
不同公司托管同一模型。OpenRouter 根据您选择的路由模式将请求路由到其中一家公司:Balanced(价格 + 速度)、Nitro(速度最快)或 Floor(价格最低)。
| US$1.60 | US$4.80 | US$0.40 | -- | -- |
- 输入 US$1.60 / 输出 US$4.80-- tok/s-- 正常运行时间
定价
客户为此模型实际支付的平均价格,与提供商公布的价格并列展示。由于缓存和折扣,实际支付的价格往往远低于标价。
加权平均值
Grok 4.7 (batch) — 价格历史
| US$0.00 | US$0.00 | 88.1% | 100.0% |
- 有效输入价格 US$0.00 / 实际输出 US$0.00token 占比100.0%
性能
吞吐量是模型生成内容的速度(每秒 token 数,越高越好)。延迟是完整的往返时间(越低越好)。TTFT 是首个 token 的响应时间,即您看到内容出现前需要等待的时间(越低越好)。
正常运行时间
正常运行时间是指过去 3 天中至少有一家提供商响应请求的时间百分比。可用性是指成功提供推理服务的时间百分比。OpenRouter 会持续监控,并在某家提供商返回错误时使用次优提供商。
正常运行时间数据尚不足,暂时无法显示。
当上游提供商发生错误时,如果您的请求筛选条件允许,我们可以通过路由到另一家健康的提供商来恢复服务。您可以通过以下接口以编程方式访问各提供商的正常运行时间数据: 端点 API。了解更多有关我们的 负载均衡和提供商路由选项。
基准测试
标准化评测得分,以及此模型针对共用基准测试提示词的实际输出。百分比越高越好,每项排名显示此模型在该来源所评测的模型中的位置。
| 来源 | 基准测试 | 得分 |
|---|---|---|
| Artificial Analysis | Grok 4.7 (High) Intelligence Index | 35.2 |
| Artificial Analysis | Grok 4.7 (High) HLE | 42.3% |
| Artificial Analysis | Grok 4.7 (High) AA-LCR | 77.0% |
| Artificial Analysis | Grok 4.7 (High) GDPval-AA | 60.9% |
| Artificial Analysis | Grok 4.7 (High) CritPt | 18.0% |
| Artificial Analysis | Grok 4.7 (High) SciCode | 57.8% |
| Artificial Analysis | Grok 4.7 (High) Terminal-Bench 4.0 | 24.7% |
| Artificial Analysis | Grok 4.7 (High) AA-Omniscience Accuracy | 41.8% |
| Artificial Analysis | Grok 4.7 (High) AA-Omniscience Non-Hallucination Rate | 79.8% |
| Artificial Analysis | Grok 4.7 (Low) Intelligence Index | 31.1 |
| Artificial Analysis | Grok 4.7 (Low) HLE | 39.2% |
| Artificial Analysis | Grok 4.7 (Low) AA-LCR | 80.0% |
| Artificial Analysis | Grok 4.7 (Low) GDPval-AA | 54.4% |
| Artificial Analysis | Grok 4.7 (Low) CritPt | 13.4% |
| Artificial Analysis | Grok 4.7 (Low) SciCode | 54.9% |
| Artificial Analysis | Grok 4.7 (Low) Terminal-Bench 4.0 | 16.2% |
| Artificial Analysis | Grok 4.7 (Low) AA-Omniscience Accuracy | 45.7% |
| Artificial Analysis | Grok 4.7 (Low) AA-Omniscience Non-Hallucination Rate | 66.2% |
| Artificial Analysis | Grok 4.7 (Xhigh) Intelligence Index | 35.5 |
| Artificial Analysis | Grok 4.7 (Xhigh) HLE | 43.1% |
| Artificial Analysis | Grok 4.7 (Xhigh) AA-LCR | 76.7% |
| Artificial Analysis | Grok 4.7 (Xhigh) GDPval-AA | 60.8% |
| Artificial Analysis | Grok 4.7 (Xhigh) CritPt | 17.7% |
| Artificial Analysis | Grok 4.7 (Xhigh) SciCode | 57.4% |
| Artificial Analysis | Grok 4.7 (Xhigh) Terminal-Bench 4.0 | 25.8% |
| Artificial Analysis | Grok 4.7 (Xhigh) AA-Omniscience Accuracy | 42.7% |
| Artificial Analysis | Grok 4.7 (Xhigh) AA-Omniscience Non-Hallucination Rate | 80.0% |
| Design Arena | Grok 4.7 Agents Arena Blender3d Elo | 1163 |
| Design Arena | Grok 4.7 Agents Arena Codetovideo Elo | 1109 |
| Design Arena | Grok 4.7 Agents Arena Mobile Apps Elo | 1242 |
| Design Arena | Grok 4.7 Agents Arena Pdf Elo | 1173 |
| Design Arena | Grok 4.7 Agents Arena Spreadsheet Elo | 1420 |
| Design Arena | Grok 4.7 Models Arena 3D Elo | 1251 |
| Design Arena | Grok 4.7 Models Arena Code Categories Elo | 1266 |
| Design Arena | Grok 4.7 Models Arena Data Visualization Elo | 1223 |
| Design Arena | Grok 4.7 Models Arena Game Development Elo | 1305 |
| Design Arena | Grok 4.7 Models Arena SVG Elo | 1271 |
| Design Arena | Grok 4.7 Models Arena UI Component Elo | 1276 |
| Design Arena | Grok 4.7 Models Arena Website Elo | 1244 |
| Vals AI | Grok 4.7 Access to Justice Benchmark | 73.7% |
| Vals AI | Grok 4.7 BioMysteryBench | 69.3% |
| Vals AI | Grok 4.7 Code Migration v1.1 | 53.2% |
| Vals AI | Grok 4.7 CyberBench v1.1 | 69.5% |
| Vals AI | Grok 4.7 CyberOffense | 42.3% |
| Vals AI | Grok 4.7 EMB v1.1 | 71.5% |
| Vals AI | Grok 4.7 Finance Agent v2.1 | 60.6% |
| Vals AI | Grok 4.7 Vals Finance Index | 58.8% |
| Vals AI | Grok 4.7 Financial Coaching Bench | 74.1% |
| Vals AI | Grok 4.7 HLAB-Vals-1.1 | 18.3% |
| Vals AI | Grok 4.7 Immigration | 68.0% |
| Vals AI | Grok 4.7 IOI | 57.7% |
| Vals AI | Grok 4.7 LegalBench | 84.4% |
| Vals AI | Grok 4.7 Vals Legal Index | 72.7% |
| Vals AI | Grok 4.7 Legal Research Bench | 47.1% |
| Vals AI | Grok 4.7 MedCode | 49.6% |
| Vals AI | Grok 4.7 MedScribe | 89.4% |
| Vals AI | Grok 4.7 MysteryMechanism | 25.2% |
| Vals AI | Grok 4.7 ProgramBench | 0.5% |
| Vals AI | Grok 4.7 ProofBench v1.1 | 26.0% |
| Vals AI | Grok 4.7 Public Benefits Bench v1.1 | 65.6% |
| Vals AI | Grok 4.7 Robo-bench | 0.2% |
| Vals AI | Grok 4.7 Vals RSI Index | 20.2% |
| Vals AI | Grok 4.7 SAGE | 88.7% |
| Vals AI | Grok 4.7 Vals Science Index | 34.8% |
| Vals AI | Grok 4.7 SRE Bench | 3.8% |
| Vals AI | Grok 4.7 SWE-bench Multimodal | 42.9% |
| Vals AI | Grok 4.7 Tax Agent Bench | 35.8% |
| Vals AI | Grok 4.7 Terminal-Bench 4.0 | 28.8% |
| Vals AI | Grok 4.7 Terminal-Bench-Science | 10.0% |
| Vals AI | Grok 4.7 Vals Index | 58.6% |
| Vals AI | Grok 4.7 Vibe Code Bench 1-100 v1.1 | 26.4% |
| Vals AI | Grok 4.7 Vibe Code Bench v1.1 | 86.2% |
应用
向此模型发送最多流量的公开应用。这能很好地反映实际生产工作负载的情况,也能提示此模型最适合哪些用例。
使用此模型的应用尚无足够数据。
活动
此模型的 token 用量和请求流量随时间的变化。
API
可直接用于调用此模型的代码。OpenRouter 的 API 与 OpenAI 兼容——大多数 SDK 只需替换基础 URL 即可使用。切换模型时,唯一需要更改的是下方的模型标识符。
目前无法通过 EU 区域内路由使用。 区域内路由文档
文档
在 OpenRouter 上使用 SpaceXAI: Grok 4.7 (batch) 构建应用的指南、实用教程、博客文章和工具。
常见问题
Grok 4.7 is SpaceXAI's flagship model for coding, agentic tasks, and knowledge work, succeeding Grok 4.6. It is particularly strong at long-running software engineering tasks, verifying its own work, and managing long context, and it improves on its predecessor at professional knowledge work such as drafting documents and presentations.
OpenRouter 上的 Grok 4.7 (batch) 文档包括 How to Benchmark LLMs 博客文章和How to Stop an AI Agent Stuck in an Infinite Loop 博客文章。快速入门介绍了 API 密钥和您的首次请求,API 参考列出了所有参数。
Grok 4.7 (batch) 的费用为 US$1.60/百万输入 token 和 US$4.80/百万输出 token,Cache Read,价格为 US$0.40/M tokens和Web Search,价格为 US$5.00/1K calls 单独计费。
Grok 4.7 (batch) 的上下文窗口为 500,000 token。
是的。Grok 4.7 (batch) 接受 tools 和 tool_choice,以进行函数调用。 它还支持通过 response_format 中的 JSON schema 实现结构化输出。
Grok 4.7 (batch) 接受 文本、图像和PDF 等文件 作为输入,并返回 文本。
在 OpenRouter 上,Grok 4.7 (batch) 由 1 家提供商提供服务:SpaceXAI。
Grok 4.6, Grok 4.5, Grok Build 0.1 及另外 3 个 是 SpaceXAI 的其他 文本 模型。
Grok 4.7 (batch) 于 2026年9月21日 发布。