A New Generation of Kimi Models in Ayeto
We have added four new options from Moonshot AI to Ayeto: Kimi K2.6, Kimi K2.7 Code, its faster Kimi K2.7 Code HighSpeed variant, and the flagship Kimi K3. Together, they cover large software projects, agentic workflows, visual understanding, and demanding knowledge work.
You can select a model for a specific AI assistant and switch it later as the task changes. Your instructions, tools, and workflow can stay in Ayeto instead of moving to another provider.
Kimi K2.6: A Multimodal Model for Long Workflows
Kimi K2.6 is an open-source, multimodal agentic model. Moonshot AI designed it for long-horizon coding, building interfaces from textual or visual input, autonomous execution, and multi-agent coordination.
It is a good fit for:
- changing a large repository across multiple files,
- building a front end from a screenshot or design,
- multi-step tasks that use tools,
- processing larger projects without repeatedly restating context.
The model supports text, image, and video input and provides a 256,000-token context window.
Kimi K2.7 Code: A Specialist for Agentic Coding
Kimi K2.7 Code focuses specifically on software development. It suits tasks where an assistant must work on code for an extended period, follow instructions accurately, reason through several steps, and repeatedly use development tools.
Ayeto also offers Kimi K2.7 Code HighSpeed. It is the same model served in a faster configuration—Moonshot AI reports approximately 180 tokens per second and up to 260 tokens per second with short contexts. Choose HighSpeed for interactive development and fast iterations; choose the standard variant when response speed is not the main criterion.
Both variants support up to 256,000 tokens of context. K2.7 Code always operates with reasoning enabled, making it better suited to complex development than simple text responses.
Kimi K3: The Flagship Model for the Most Demanding Tasks
Kimi K3 is Moonshot AI’s most capable flagship model. It combines long-horizon coding, knowledge work, advanced reasoning, native visual understanding, and tool use. Its context window of up to 1 million tokens lets it work with extensive source material or a large portion of a codebase in one request.
Choose K3 when you need to:
- analyze a complex problem with extensive source material,
- work on large software projects over long horizons,
- combine code, documents, and visual input,
- manage a complex agentic workflow with tools.
Reasoning is always enabled, and its effort can be configured. That means K3 is not automatically the best choice for every short or cost-sensitive task—the model should match the work.
Which Model Should You Choose?
| Need | Recommended model |
|---|---|
| General multimodal and agentic work | Kimi K2.6 |
| Long-horizon coding and tool use | Kimi K2.7 Code |
| Fast, interactive coding iterations | Kimi K2.7 Code HighSpeed |
| The most demanding coding, reasoning, and knowledge work | Kimi K3 |
| Exceptionally large context | Kimi K3 |
If you are unsure, send the same prompt to several models using model comparison in Ayeto. You can compare their answers, processing time, and credit cost side by side.
Try the New Kimi Models
Open your assistant settings in Ayeto and select a model based on the task. Start with K2.6 for broad multimodal workflows, use K2.7 Code or HighSpeed for development, and choose K3 for the most demanding assignments.
Try the new Kimi models in Ayeto and compare them on a task you genuinely need to solve—not just a demo prompt.