Grok 4.6vsGrok 4.5
在 10 个共同 benchmark 中，Grok 4.6 整体领先：Grok 4.6 领先 10 项，Grok 4.5 领先 0 项，持平 0 项，平均分差 +54.32。
Grok 4.610 项(100%)(0%)0 项Grok 4.5
评测分数
按能力类目分组，每组内按分差大小排列；共 10 项。
Coding and Software Engineer
Grok 4.6 领先 4/4| 评测项 | Grok 4.6 | Grok 4.5 | 分差 |
|---|---|---|---|
| DeepSWE | 65.909 / 26Thinking High (With Tools) | 5317 / 26Thinking High (With Tools) | +12.90 |
| FrontierCode 1.1 | 61.302 / 5Thinking High (With Tools) | 56.604 / 5Thinking High (With Tools) | +4.70 |
| CursorBench 3.2 | 69.902 / 4Thinking High (With Tools) | 66.704 / 4Thinking High (With Tools) | +3.20 |
| APEX-SWE | 56.402 / 3Thinking High (With Tools) | 53.603 / 3Thinking High (With Tools) | +2.80 |
Productivity Knowledge
Grok 4.6 领先 3/3| 评测项 | Grok 4.6 | Grok 4.5 | 分差 |
|---|---|---|---|
| AA-Briefcase | 1,5772 / 6Thinking High (With Tools) | 1,3136 / 6Thinking High (With Tools) | +264 |
| GDPval-AA v2 | 1,7533 / 13Thinking High (With Tools) | 1,5267 / 13Thinking High (With Tools) | +227 |
| Harvey Lab-AA | 15.803 / 6Thinking High (With Tools) | 12.904 / 6Thinking High (With Tools) | +2.90 |
Agent Level Benchmark
Grok 4.6 领先 1/1| 评测项 | Grok 4.6 | Grok 4.5 | 分差 |
|---|---|---|---|
| APEX-Agents | 57.502 / 5Thinking High (With Tools) | 47.104 / 5Thinking High (With Tools) | +10.40 |
AI Agent - Tool Usage
Grok 4.6 领先 1/1| 评测项 | Grok 4.6 | Grok 4.5 | 分差 |
|---|---|---|---|
| Terminal-Bench 3.0 | 264 / 6Thinking High (With Tools) | 15.705 / 6Thinking High (With Tools) | +10.30 |
General Knowledge
Grok 4.6 领先 1/1| 评测项 | Grok 4.6 | Grok 4.5 | 分差 |
|---|---|---|---|
| AA Intelligence Index | 612 / 8Thinking High (With Tools) | 564 / 8Thinking High (With Tools) | +5 |
规格对比
| 字段 | Grok 4.6 | Grok 4.5 |
|---|---|---|
| 发布机构 | xAI | xAI |
| 发布时间 | 2026-08-12 | 2026-07-08 |
| 模型类型 | 编程大模型 | 编程大模型 |
| 架构 | 稠密模型 | 稠密模型 |
| 参数规模 | 暂无数据 | 暂无数据 |
| 上下文长度 | 500K | 500K |
| 最大输出 | 暂无数据 | 暂无数据 |
API 调用价格
价格优先使用 DataLearner 配置的 API 记录；缺失项不做推测。
| 价格项 | Grok 4.6 | Grok 4.5 |
|---|---|---|
| 文本输入 | $2 / 1M tokens | $2 / 1M tokens |
| 文本输出 | $6 / 1M tokens | $6 / 1M tokens |
| 缓存读取 | $0.5 / 1M tokens | $0.5 / 1M tokens |
小结
- Grok 4.6在以下类目领先:Coding and Software Engineer (4/4)、Productivity Knowledge (3/3)、Agent Level Benchmark (1/1)、AI Agent - Tool Usage (1/1)、General Knowledge (1/1)
10 个共同 benchmark 上，Grok 4.6 平均高出 54.32 分。
单项差距最大的 benchmark：AA-Briefcase — Grok 4.6 1,577，Grok 4.5 1,313（分差 +264）。
本页正文由结构化模型、价格与 benchmark 数据生成，不使用实时 LLM 撰写。