Kimi K2.7-Code: open-source coding model with better token efficiency
huggingface.co
June 12, 2026
9 min read
🔥🔥🔥🔥🔥
66/100
Summary
moonshotai/Kimi-K2.7-Code can be utilized with libraries like Transformers for image-text-to-text tasks. A high-level pipeline can be created using the command `pipe = pipeline("image-text-to-text", model="moonshotai/Kimi-K2.7-Code", trust_remote_code=True)`.
Key Takeaways
- Kimi K2.7 Code is a coding-focused model that improves long-horizon coding task completion and reduces token usage by approximately 30% compared to its predecessor, Kimi K2.6.
- The model architecture is based on a Mixture-of-Experts (MoE) design, featuring 1 trillion total parameters and 384 experts.
- Users can utilize Kimi K2.7 Code through various platforms including Transformers, vLLM, SGLang, and Docker.
- The model supports image-text interactions, allowing users to input images and receive descriptive text outputs.
Coverage & discussions
Hacker News · huggingface.co
Community discussions
Community Sentiment
Positives
- Kimi K2.7's improved token efficiency could significantly reduce costs for developers, making advanced AI more accessible for personal projects.
- Users appreciate Kimi models for their usability in coding tasks, indicating they can be effective alternatives to pricier models like Claude.
- The competitive pricing of Kimi models suggests a potential shift in the market, encouraging innovation and democratization of AI tools.
Concerns
- Despite its lower cost, Kimi K2.6 is reportedly outperformed by models like Claude Sonnet 4.6 and GPT 5.4 Mini, raising concerns about its competitiveness.
- Some users express frustration that Kimi models struggle with complex cognitive tasks, which limits their effectiveness compared to higher-end models.
- There are doubts about the true performance gap between Kimi and more expensive models, with some users feeling the differences are not as pronounced as advertised.
