Kimi K2.6
Kimi K2.6 is a large language model released by Moonshot AI and part of the Kimi K model series. The model features code generation and front-end programming capabilities, supporting a range of programming task scenarios. As an AI model oriented toward developers, Kimi K2.6 can be integrated via API and is applicable to use cases including coding assistance and content generation.
Core Facts
Timeline (last 90 days)
Kimi K2.6 achieved a success rate of 7.5% on CAD World, OpenCUA achieved 1.5%, and two other models achieved a success rate of 0.
kimi-2-6-code is powered under the hood by Moonshot's Kimi K2.6 model.
The Kimi K2 series supports a 128K token context.
On BU Bench V1, the 4B-parameter SFT model surpassed the 45% score reported by DeepSeek V4 Pro and Kimi K2.6 under the open-source Browser Use framework.
All Facts (10)
Kimi K2.6 supports an ultra-long context window of 256K.
80%VerifiedKimi K2.6 adopts a Mixture of Experts (MoE) architecture, with a total parameter count at the trillion level and only 32B parameters actually activated.
80%UnverifiedK2.6采用Modified MIT License许可证
85%UnverifiedKimi K2.6能根据问题复杂度自适应调整输出长度，这一能力通过Post Training阶段的长度校准技术实现
80%UnverifiedKimi K2.6支持300个并行子代理进行4000步的协作
80%UnverifiedSGLang对MoE模型的专家并行支持比vLLM更为成熟，使其成为部署Mixtral、DeepSeek、Kimi K2.6等MoE架构模型的首选框架之一
70%UnverifiedKimi K2.6 achieved a success rate of 7.5% on CAD World, OpenCUA achieved 1.5%, and two other models achieved a success rate of 0.
50%Unverifiedkimi-2-6-code is powered under the hood by Moonshot's Kimi K2.6 model.
50%UnverifiedThe Kimi K2 series supports a 128K token context.
50%UnverifiedOn BU Bench V1, the 4B-parameter SFT model surpassed the 45% score reported by DeepSeek V4 Pro and Kimi K2.6 under the open-source Browser Use framework.
50%