Moonshot 发布 Kimi K3.1 Pro，发布前削弱其漏洞利用能力Chinese only
Moonshot 称，在训练中移除网络安全相关数据后，Kimi K3.1 Pro 在其内部漏洞利用测试中的通过率从未过滤版本的 90.5% 降至 81.0%，仍高于 K3 的 33.3%。该模型及速度快 2—3 倍的 Highspeed 版本已在 Kimi 应用和 API 上线；权重计划于 2026 年 11 月 4 日前开放。
1 developments4 reports3 sources2 verified facts
Verified facts 2
IT早报 1024：月之暗面发布 Kimi K3.1 Pro；诋毁嘲讽尊界 S800 自媒体博主向尊界道歉；华为 Pura90 Plus 开启预订；iPhone Duo 内屏膜撕下秒变亮面屏...
黄仁勋表示对英伟达与长鑫存储合作持开放态度，并称中国在基础设施技术和制造业方面能力很强，应认真对待。
IT早报 1024：月之暗面发布 Kimi K3.1 Pro；诋毁嘲讽尊界 S800 自媒体博主向尊界道歉；华为 Pura90 Plus 开启预订；iPhone Duo 内屏膜撕下秒变亮面屏...
苹果盘中总市值暂时超越英伟达，使英伟达暂时失去全球上市公司市值榜首位置。
Event developments 1
Moonshot releases Kimi K3.1 Pro after reducing its exploit capability
Moonshot says removing cyber-related training data lowered Kimi K3.1 Pro’s pass rate on its internal exploit test from 90.5% for the unfiltered version to 81.0%, still above K3’s 33.3%. The model and a Highspeed variant running 2–3 times faster are live in Kimi apps and the API; weights are due by November 4, 2026.
View source reports · 3 sources · 4 reports
- Rohan Paul (@rohanpaul_ai) · Moonshot releases Kimi K3.1 Pro after reducing its exploit capability
- MarkTechPost AI 研究报道 · Moonshot AI releases Kimi K3.1 Pro via API; weights expected in November
- IT之家 科技新闻 · Moonshot AI releases Kimi K3.1 Pro with a one-million-token context window and 3D Make
- IT之家 科技新闻 · Moonshot AI releases Kimi K3.1 Pro with support for a 1 million context window
All reports 4
Moonshot AI releases Kimi K3.1 Pro with support for a 1 million context window
Moonshot AI's Kimi announced K3.1 Pro as the first model in its Kimi K3.1 series. It describes the model as its most capable to date, with parameters in the 3 trillion range and support for a 1 million context window. The material gives no benchmark results or access details.
3 more sources reported this
- IT之家 科技新闻
- MarkTechPost AI 研究报道
- Rohan Paul (@rohanpaul_ai)
Moonshot AI releases Kimi K3.1 Pro with a one-million-token context window and 3D Make
Kimi K3.1 Pro is available through Kimi’s website, app, Work, Code and API; full model weights are due by November 4. Users can preview and edit models in 3D Make, then download them for 3D printing. According to the company’s description, the model scored 65.7% on the Terminal-Bench 4.0 coding-agent benchmark.
3 more sources reported this
- IT之家 科技新闻
- MarkTechPost AI 研究报道
- Rohan Paul (@rohanpaul_ai)
Moonshot AI releases Kimi K3.1 Pro via API; weights expected in November
Kimi K3.1 Pro is live on Kimi.ai, Kimi Work, Kimi Code and the API. It accepts text, images and video and has a 1-million-token context window. API pricing is $2 per million input tokens and $10 per million output tokens. Weights are not public yet and are expected in November 2026. Moonshot's self-reported results show strong coding performance but a substantial gap on scientific terminal tasks against the competitors in its chart.
2 more sources reported this
- IT之家 科技新闻
- Rohan Paul (@rohanpaul_ai)
Moonshot releases Kimi K3.1 Pro after reducing its exploit capability
Moonshot says removing cyber-related training data lowered Kimi K3.1 Pro’s pass rate on its internal exploit test from 90.5% for the unfiltered version to 81.0%, still above K3’s 33.3%. The model and a Highspeed variant running 2–3 times faster are live in Kimi apps and the API; weights are due by November 4, 2026.
2 more sources reported this
- IT之家 科技新闻
- MarkTechPost AI 研究报道