# Grok 4.7 发布：Harvey 法律基准 19.6% 大幅领先，Terminal-Bench 4.0 得分近翻倍且价格不变

- 来源：Rohan Paul (@rohanpaul_ai)
- 发布时间：2026-09-22 02:47
- AIHOT 分数：68
- AIHOT 链接：https://aihot.news/items/cmubmgvut049lro0lbsamrioy
- 原文链接：https://x.com/rohanpaul_ai/status/2102107446852592045

## AI 摘要

Grok 4.7 发布，官方称在同等价格和速度下较 Grok 4.6 有显著提升。Harvey Legal Agent Benchmark 上 Grok 4.7 得分 19.6%。

## 正文

Legal work saw one of Grok 4.7’s biggest jumps.

On the Harvey Legal Agent Benchmark, Grok 4.7 scored 19.6%, while GPT-5.6 Sol Max scored 2.5% and Fable 5.1 Max scored 6.7%

Also, Long-running terminal work nearly doubled. Grok 4.7 reached 38.0% on Terminal-Bench 4.0, versus 20.3% for Grok 4.6, suggesting the larger improvement is in agents that must keep working through extended computer tasks.

And SpaceXAI made those gains without raising the headline token price. Grok 4.7 remains at $2/$6 per 1M input/output tokens.

### 引用推文

> SpaceXAI：Grok 4.7 is here. It's a notable improvement over Grok 4.6 at the same price and speed.