Moonshot AI’s Hugging Face model card describes Kimi K2.7 Code as a 1T-parameter mixture-of-experts coding model with 32B active parameters and a 256K context length, while Kimi’s API and product pages list availability, pricing, and CLI integration.
Moonshot AI has published a Hugging Face model card for Kimi K2.7 Code, presenting it as a coding-focused model for agentic software-development workflows.
In the Hugging Face model card for moonshotai/Kimi-K2.7-Code, Moonshot AI describes Kimi K2.7 Code as a mixture-of-experts model with 1T total parameters and 32B active parameters. The same model card says the model supports a 256K context length, positioning it for tasks that may involve large codebases or long development traces.
Moonshot AI’s Hugging Face page also includes OpenAI-compatible deployment examples, according to the source document. That detail matters for developers because OpenAI-compatible interfaces can reduce integration work when teams are already using tools or applications built around similar API patterns.
The model card further includes benchmark comparisons against Kimi K2.6, GPT-5.5, and Claude Opus 4.8. The available source excerpt does not provide the underlying benchmark scores or methodology, so those comparisons should be read as Moonshot AI’s published benchmark framing rather than independently verified performance claims.
Moonshot AI’s Kimi API Platform says Kimi K2.7 Code has been officially released and lists it under “Latest Models.” The platform describes it as Kimi’s strongest coding model and shows pricing of $0.95 per million input tokens and $4.00 per million output tokens.
Those public prices give developers a baseline for evaluating the model against other code-oriented large language models. However, the cited API page excerpt does not specify details such as rate limits, regional availability, enterprise terms, or whether discounts apply at higher usage levels.
Moonshot AI’s Kimi Code product page says K2.7 Code has been updated and presents the Kimi Code CLI as powered by K2.7 Code for software-development workflows. The page describes the product in the context of an AI code agent and command-line tooling, suggesting that Moonshot AI is pairing the model release with developer-facing interfaces rather than offering it only as a hosted API endpoint.
The combination of a Hugging Face model card, API platform listing, and Kimi Code product page gives K2.7 Code three public entry points: model documentation, hosted access, and a CLI-oriented developer product. Each source emphasizes coding use cases, but the available excerpts do not include independent evaluations, user adoption figures, or detailed safety documentation.
Kimi K2.7 Code arrives as model providers continue to compete on long-context coding, tool use, and developer workflow integration. Moonshot AI is emphasizing model scale, a 256K context window, OpenAI-compatible deployment examples, and a dedicated CLI product. For engineering teams, the most concrete public details are the model architecture summary, context length, API pricing, and Kimi Code integration described by Moonshot AI’s own materials.
As with any vendor-published model release, developers should review the full model card, test the model on their own repositories and tasks, and compare cost, latency, reliability, and licensing requirements before adopting it in production.
Moonshot AI has published a Hugging Face model card for Kimi K2.7 Code, presenting it as a coding focused model for agentic software development workflows.
What Moonshot AI disclosed In the Hugging Face model card for moonshotai/Kimi K2.7 Code , Moonshot AI describes Kimi K2.7 Code as a mixture of experts model with 1T total parameters and 32B active parameters.
The same model card says the model supports a 256K context length, positioning it for tasks that may involve large codebases or long development traces.
Continue reading
