人工分析,2026 年 9 月 2 日。最大变体得分为 52%,是其列出的所有模型中得分最高的,但未绘制图表。 · 来源
| 规格 | 价值 |
|---|---|
| 创客 | 元 |
| 已发布 | 2026 年 9 月 2 日 |
| 上下文窗口 | 100 万个代币 |
| 输入 | 视频、图像和文档(元);文本、图像和视频(人工分析) |
| API模型 | muse-spark-1.3 和 muse-spark-1.3-contributor |
| muse-spark-1.3 每 100 万个代币的价格(美元) | $1.25 输入,$0.15 缓存输入,$4.25 输出 |
| 每 100 万个代币的 muse-spark-1.3-contributor 价格(美元) | $0.10 输入,$0.002 缓存输入,$0.20 输出;用于改进Meta[1][2]的产品 |
| 每个智力指数任务的成本(xhigh) | Muse Spark 1.2 为 0.55 美元,而 Muse Spark 1.2 为 0.40 美元 |
| 使用权 | Muse 代码和元模型 API |
来自 Meta 的开发者模型页面(于 2026 年 11 月 2 日开放)和 Artificial Analysis 的发布文章。The page does not state a maximum output length. · 来源
Estimate from the version's published specs, benchmarks and reviews, 0-5; scores are only comparable within one criteria year. September 2 launch, generally available xhigh variant: Artificial Analysis launch Index 61 (contemporary index), 1M context, text/image/video. Standard API USD1.25 input/4.25 output per million gives USD2 blended at 3:1. Limited-preview max 62 is not the shipping variant. Terminal-Bench 2.1 85% cannot populate the Terminal-Bench 4 coding axis. Launch throughput not established, so speed null. Current rebased Index values and later speed are not substituted. DataCamp tested six runs on three coding tasks; IonWarp is a review-tool vendor reporting its own benchmark, not an independent laboratory.[2]
Both variants' gains come primarily from improvements in agentic work and scientific capabilities
Across 6 runs on 3 coding tasks, 1.3 was cheaper on 2 and 38% more expensive on the third, for a net 12% cost increase over the set.
Muse Spark 1.3 is the best model we have run on general code review.
| Muse Spark 1.3: Meta reaches the frontier · 2026年9月2日 | 0.50 |
|---|---|
| Muse Spark 1.3 Tutorial: A Hands-On Developer Guide · 2026年9月7日 | -0.20 |
| Muse Spark code review: Muse Spark 1.3 benchmarks on real pull requests · 2026年10月5日 | 0.80 |
读者上传的照片,与图钉带来源的媒体分开显示。你可以为吃到或看到的内容评分,在餐厅图钉上还能直接在照片上标记菜品。
第一项始终是图钉的来源。总体可信度是各份资料对上方所用开始和结束时间支持程度的加权平均;资料每比最新的一份旧 180 天,权重减半。
Meta[2] 的博文日期为“2026 年 9 月 2 日”,文中写道:“我们很高兴发布 Muse Spark 1.3”,并称其“今天已在 Muse Code 和 Meta Model API 中上线”;Artificial Analysis[4] 同日发布的文章开篇即写道:“Meta 已发布 Muse Spark 1.3”。
Meta's[1] developer model page lists the two API models and their prices - muse-spark-1.3 at $1.25 input, $0.15 cached input and $4.25 output per million tokens, and muse-spark-1.3-contributor at $0.10 and $0.20 with data used to improve Meta's products - both with a 1M-token context window.
VentureBeat's report by Carl Franzen says the xhigh version is broadly available through Muse Code and Meta[1][2] Model API while the max configuration is undergoing safety testing and will arrive shortly, and notes the open-weights release now on Meta's roadmap.
Artificial Analysis' same-day independent benchmark write-up: Muse Spark 1.3 (xhigh) scores 61 on its Intelligence Index, up from 57 for 1.2 and 53 for 1.1, and the limited-preview max variant 62, behind only Claude Fable 5.1 and Claude Opus 5; it gives the Tau3-Bench Banking, Terminal-Bench 2.1, GDPval-AA v2 and cost-per-task figures.
有遗漏或错误吗?用你自己的话说明:能佐证此图钉的链接、不同的开始或结束日期及理由,或缺失、有误的信息。AI 会对照此图钉的来源进行核实,搜索更好的来源,并添加任何支持你说法的页面。图钉自身的来源仍然最重要。AI 也会查看图片:显示的是别的东西或显示效果差的图片会被移到后面或替换。
原标题: Muse Spark 1.3 Released