# Meta 发布 Muse Spark 1.3，长上下文 MRCR 256K-512K 得分 98.5

- 来源：Rohan Paul (@rohanpaul_ai)
- 发布时间：2026-09-03 03:49
- AIHOT 分数：65
- AIHOT 链接：https://aihot.virxact.com/items/cmtkj69ou037groll0nd8fp16
- 原文链接：https://x.com/rohanpaul_ai/status/2095237722046881912

## AI 摘要

Meta 发布 Muse Spark 1.3，官方称比 1.2 减少 20% 工具调用和 25% token 用量。作者汇总基准数据：MRCR 256K–512K 得分 98.5，超过 GPT-5.6 Sol 的 91.5；JobBench 64.9、OSWorld 66.9、AutomationBench 49.4，接近 Opus 5 的 65.7、68.3、50.3。

## 正文

Meta released Muse Spark 1.3. Reports 20% fewer tool calls and 25% fewer tokens than 1.2.

So long coding jobs should require fewer model actions and less generated text to reach the same deliverable.

• Long context is the standout: Muse Spark 1.3 scores 98.5 on MRCR 256K–512K. Beats GPT-5.6 Sol

• it gets very close to Opus 5 on JobBench (64.9 vs 65.7), OSWorld (66.9 vs 68.3) and AutomationBench (49.4 vs 50.3)

### 引用推文

> Mark Zuckerberg：Muse Spark 1.3 is rolling out today with frontier performance almost too cheap to meter. This is the biggest jump we've made so far on coding and agentic work. ...