快手技术团队｜KStech

Post the latest developments and achievements of Kuaishou technology.

[快手技术团队｜KStech @KStech0606]

Website:https://t.co/IzWzQboUEx
Submission Site:https://t.co/8TRRkweuXE

[快手技术团队｜KStech @KStech0606]

FRecSys 2026 @ ACM SIGOPS ATC 2026 is now inviting paper submissions.

We cordially invite contributions from a broad range of research communities, including (but not limited to) Recommender Systems, AI Infrastructure, ML Systems, Data Infrastructure, Databases, and Networking.

[快手技术团队｜KStech @KStech0606]

征集令｜快手探索者LLM-Rec挑战赛正式发布！

[KwaiKeye @KwaiKeye]

🚀 Keye-VL-2.0-30B-A3B released — fully open.

The first VLM to land DSA (DeepSeek Sparse Attention) in production — nearly lossless reasoning over 256K ultra-long context.

30B params. Fully open. 🤗

🧵👇

[KwaiKeye @KwaiKeye]

Find us on Hugging Face & GitHub 👇
Hugging Face： https://t.co/3sHrRtMETe
GitHub：https://t.co/huZEOklhB8

[快手技术团队｜KStech @KStech0606]

Huge congrats to the @KwaiAICoder team on KAT‑Coder‑Pro V2’s new results on the @ArtificialAnlys Intelligence Index — now the top non‑reasoning model in their benchmarks!

[Artificial Analysis @ArtificialAnlys]

KwaiKAT has released KAT-Coder-Pro V2, a non-reasoning model that scores 44 on the Artificial Analysis Intelligence Index, an 8 point improvement from KAT-Coder-Pro V1

@KwaiAICoder has updated their flagship proprietary coding model with the release of KAT-Coder-Pro V2. KAT-Coder-Pro V2 achieves 44 on the Artificial Analysis Intelligence Index, matching Claude Sonnet 4.6 (non-reasoning) and trailing only Claude Opus 4.6 (non-reasoning, 46) among non-reasoning models. At ~9M output tokens, it is also more token efficient than Claude Opus 4.6 (~11M), Claude Sonnet 4.6 (~14M), and reasoning models with similar intelligence such as DeepSeek V3.2 (reasoning, ~61M) and Qwen3.5 397B A17B (reasoning, ~86M).

KAT-Coder-Pro V2 is a non-reasoning model, unlike all of the current frontier language models which ‘think’ before answering. Typically, reasoning variants score higher on the Intelligence Index than their non-reasoning counterparts, but consume more output tokens and are less suited to latency-sensitive workloads.

Key Highlights:

➤ 🧠 Higher overall intelligence, but regression in long context reasoning and knowledge recall: KAT-Coder-Pro V2 scores 44 on the Artificial Analysis Intelligence Index, an 8 point improvement from KAT-Coder-Pro V1 and matching Claude Sonnet 4.6 (non-reasoning, max effort). It performs well on tool use (90% on Tau2-Telecom), but regresses compared to KAT-Coder-Pro V1 on long-context reasoning and knowledge, falling 8 p.p. on AA-LCR (66%) and 17 p.p. on HLE (16%).

➤ 🤖 Agentic capability improvements: KAT-Coder-Pro V2 shows major improvements on our agentic evaluations. On Terminal-Bench Hard, it scores 49%, up 40 p.p. from KAT-Coder-Pro V1, making it the highest-scoring non-reasoning model, matching Claude Opus 4.6 (non-reasoning, 49%) and ahead of Claude Sonnet 4.6 (non-reasoning, 46%). KAT-Coder-Pro V2 also shows improvement in GDPval-AA, scoring 1123 (+304 Elo from V1), but still sits behind models such as DeepSeek V3.2 (1198) and Qwen3.5 397B A17B (1202).

➤ ⚙️ High token efficiency: KAT-Coder-Pro V2 is a non-reasoning model and uses fewer tokens than peers with similar intelligence. It uses 8.7M output tokens to run the Artificial Analysis Intelligence Index, below Claude Opus 4.6 (non-reasoning, ~11M) and Claude Sonnet 4.6 (non-reasoning, ~14M), though this is ~2x higher than its predecessor, KAT-Coder-Pro V1 (~4.5M). It also uses significantly fewer tokens than similarly intelligent reasoning models such as DeepSeek V3.2 (reasoning, ~61M) and Qwen3.5 397B A17B (reasoning, ~86M).

➤ $ Improved cost efficiency: KAT-Coder-Pro V2 costs $73 to run the Artificial Analysis Intelligence Index, down from $76 for V1, as it uses fewer input tokens by requiring fewer turns in agentic evaluations. This makes it one of the most cost-efficient models at its intelligence level, costing less than Qwen3.5 397B A17B (reasoning, $418) and Claude Sonnet 4.6 (non-reasoning, $1397). KAT-Coder-Pro V2 is currently priced at $0.30/$1.20 per 1M input/output tokens on StreamLake and AtlasCloud API endpoints.

➤ ⚡ Low end-to-end response time: KAT-Coder-Pro V2 runs at ~109 output tokens per second, far ahead of Claude Opus 4.6 (non-reasoning, 39 OTPS) and Claude Sonnet 4.6 (non-reasoning, 43 OTPS). Because it also has a low time to first token without any reasoning delay, it delivers one of the fastest end-to-end response times, which measures the time taken from request sent to final output returned.

Model details:

➤ Availability: KAT-Coder-Pro V2 is available via StreamLake and AtlasCloud API endpoints
➤ Context Window: 256K tokens (equivalent to KAT-Coder-Pro V1)
➤ Multi-modal capabilities: Text input and output only

[KwaiKAT @KwaiAICoder]

https://t.co/pE1zyPnCgu

[KwaiKAT @KwaiAICoder]

KAT-Coder-Pro V2 is available now - our most capable agentic coding model yet.

- Native OpenClaw support with cross-scaffold generalization across 10+ frameworks
- Breakthrough web aesthetics generation— production-grade Landing Pages & PPTs from a single sentence

Coding Plan：
https://t.co/N9Ld6MfQjw
API KEY：
https://t.co/Gu7atvvo7Q
Tool Integration Guide:
https://t.co/QLBLackoRA

[快手技术团队｜KStech @KStech0606]

不同业务视觉不统一、Vue 和 React 组件不能复用、跨团队沟通成本拉满…多业务线前端 / 设计真的太容易陷入 “重复劳动” 陷阱了😫
我们决定从真实场景出发，构建一套完全贴合快手需求、并能伴随业务共同演进的设计与研发体系——KwaiDesign 由此诞生！

[快手技术团队｜KStech @KStech0606]

快手搜索团队提出全新的检索数据引擎 CroPS。该方法通过引入用户换 Query 数据、推荐流数据以及大模型生成的世界知识，多视角丰富了正样本信号，并结合层次化标签分配（HLA）策略和 H-InfoNCE 损失函数，实现了对相关性的精细化建模，该论文已被人工智能顶级会议 AAAI 2026 Oral 接收！

[KwaiKAT @KwaiAICoder]

KAT-Coder-Pro V1： HUGE UPDATE!

The latest version of KAT-Coder-Pro V1 hits 64 on the Artificial Analysis Intelligence Index, ranking 10th globally among all models

- Boosted benchmark performance
- Strengthened reasoning capabilities
- Enhanced front-end capabilities
- Elevated tool-use precision
- Reduced hallucination

(1/n)

[快手技术团队｜KStech @KStech0606]

快手可图联合清华提出的VQRAE技术，直接打造了首个能同时产生连续+离散表征的统一tokenizer，彻底打破两者的“语言隔阂”！

核心亮点速码：
✅ 单编码器架构超高效
✅ 业内首个高维全利用率码本
✅ 两阶段训练平衡满分

[快手技术团队｜KStech @KStech0606]

谁说看懂世界的AI不会创造世界？快手可图团队联合华中科技大学搞出的VGT技术，直接把这个潜力激活了——不用从头训练大模型，只要简单“专项微调”，Qwen2.5-VL、InternVL3这些“理解派”AI，立马就能变身顶尖生图高手！

[快手技术团队｜KStech @KStech0606]

换配音总对口型失败、数字人直播几小时就脸崩、遮挡场景下编辑口型直接崩效果？快手 OmniSync 直接用三大创新把这些痛点全攻克，重新定义了口型编辑的行业标准！
✅ 无 mask 训练范式，彻底解耦口型与身份
✅ 流匹配推理，实现无限时长不漂移
✅ 动态时空 CFG，精准又不毁纹理

[快手技术团队｜KStech @KStech0606]

2.统一多级对齐的生成式推荐大模型：Align³GR
论文题目：Align³GR：Unified Multi-Level Alignment for LLM-based Generative Recommendation
论文链接：https://t.co/bbKNgdmPBD

[快手技术团队｜KStech @KStech0606]

短视频广告推荐中常存在用户真实需求与广告内容、历史点击行为脱节的 “错位” 问题。快手推出的 Align³GR 广告生成式推荐大模型，以 “对齐” 技术为核心破解该难题，通过从基础对齐工具 DAS 到进阶大模型 Align³GR 的两步关键进化，让推荐系统同时读懂广告内容、把握用户真实偏好并契合平台与商家诉求。

[快手技术团队｜KStech @KStech0606]

1.一段式量化+对齐的语义ID框架：DAS
论文标题：DAS：Dual-Aligned Semantic IDs Empowered Industrial Recommender System
论文链接：https://t.co/cIusuXmfHS

[快手技术团队｜KStech @KStech0606]

🧠 核心创新：像人一样“分步推理”
研究团队提出了一个基于情感线索引导的树状推理框架。VidEmo不再试图一步到位，而是模拟人类的认知过程，分三个阶段层层递进：
🌟感知基础属性（如眉眼姿态、头部角度）
🌟分析表情信号（如微表情、动作单元）
🌟综合推断高层情感（如“隐忍的悲伤”、“克制的喜悦”）

[快手技术团队｜KStech @KStech0606]

你是否希望AI不仅能“看见”画面，更能“读懂”人心？传统方法在分析视频情感时，往往只能进行基础分类，难以理解复杂情感的动态变化。

针对这一挑战，快手可灵团队与南开大学提出了创新解决方案——VidEmo，让AI首次实现“情智兼备”的情感推理。

[快手技术团队｜KStech @KStech0606]

论文链接：https://t.co/chatLfz9Eb
数据集：https://t.co/aaC1s5Tj8n

[快手技术团队｜KStech @KStech0606]

同个模型 Python 修 bug 超神，Java 写功能就拉胯，Web 开发溜得很，碰基础设施代码直接懵…——这些开发者社区的常见吐槽，折射出现有代码大模型评估体系的局限。快手×南大联手推出SWE-Compass堪称代码模型测评界的“神仙基准”。

这 “评估指南针” 有多牛？​
覆盖 8 大任务
8 大场景全包
10 种语言兼容