2 min read
|
Saved February 14, 2026
|
Copied!
Do you care about this?
This article discusses the latest features of Kimi's AI tools, including Kimi K2 and Kimi-Researcher. It highlights their capabilities in agentic tasks, coding, and multi-turn search, along with details about API pricing and model benchmarks.
If you do, here's more
Kimi has launched a new product called Kimi Agentic Slides, featuring a tool called Nano Banana Pro. Users can access it for free for 48 hours, allowing them to create fully editable presentations from various file types, including PDFs and images. The tool claims to offer designer-level visuals, such as infographics and illustrations, making it suitable for professional use. This service is part of Kimiโs broader strategy to enhance productivity and creativity in presentations.
Another innovation highlighted is Kimi K2, an open-source agentic model that boasts a total of 1 trillion parameters and 32 billion active mixture of experts (MoE). It has achieved state-of-the-art performance on several benchmarks, particularly in coding and agentic tasks. The pricing for using Kimi K2 via API is structured at $0.15 per million input tokens for cache hits and $2.50 for output tokens. This model is positioned as a more accessible option for developers looking to integrate advanced agentic intelligence into their applications.
Kimi-Researcher is also part of this suite, functioning as an autonomous agent designed for complex search and reasoning tasks. It scored 26.9% on Humanity's Last Exam and 69% on xbench, indicating its potential effectiveness in real-world applications. It operates with an average of 23 reasoning steps per task and explores over 200 URLs, utilizing tools like search and code for multi-turn interactions. Beta access is available, and there's a plan to open-source its pretrained and reinforcement-learned models, fostering further research and development in this area.
Questions about this article
No questions yet.