Kimi 2.6 has been released
The Chinese LLM now processes 2 million tokens and can search uploaded files, challenging GPT-4o.
Moonshot AI has launched Kimi 2.6, a significant update to its flagship large language model. The headline feature is a massive expansion of its context window to 2 million tokens, doubling the previous 1 million token limit. This allows the model to process and reason over extremely long documents, such as full-length novels, lengthy legal contracts, or extensive research papers, in a single session. The update solidifies Kimi's position as one of the leading models for long-context understanding, directly competing with offerings like Anthropic's Claude 3 and OpenAI's GPT-4.
Alongside the context boost, Kimi 2.6 introduces a practical new 'File Search' capability. Users can now upload various document formats—including PDFs, Word docs, PowerPoint presentations, Excel sheets, and plain text files—and then ask specific questions about the content. The model can locate relevant information, summarize sections, or extract key data points, functioning as an intelligent research assistant. This feature enhances Kimi's utility for professionals, students, and researchers who need to analyze dense documents quickly. The model remains free to use via its web and mobile apps, maintaining its accessibility while adding enterprise-grade functionality.
- Context window doubled to 2 million tokens, enabling analysis of book-length documents.
- New 'File Search' feature allows Q&A on uploaded PDFs, Word, Excel, and PPT files.
- Remains free to use, positioning it as a powerful, accessible alternative to paid Western models.
Why It Matters
Provides a free, high-capacity AI tool for Chinese document analysis, challenging costly enterprise solutions.