China’s Coding Model, Kimi K2.7 Code, is 6x Cheaper Than Claude. It also Grades Its Own Homework

2026-06-17 · Source: Towards AI - Medium · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Advanced, quick

Summary

Moonshot AI released Kimi K2.7 Code on June 12, 2026, an open-weight, coding-specialized model priced at \$0.95 per million input tokens and \$4.00 per million output tokens. This makes it approximately six times cheaper than Claude Opus 4.8, which costs \$5 and \$25 for input and output respectively. Kimi K2.7 Code features a 256K context window and a trillion parameters using a mixture-of-experts architecture. However, all benchmarks published for K2.7 Code, including Kimi Code Bench v2 and Program Bench, are proprietary and lack independent verification on public leaderboards like SWE-bench Verified. Furthermore, on its own in-house scorecards, K2.7 Code loses to GPT-5.5 on all six metrics and to Claude Opus 4.8 on five of six.

Key takeaway

For AI Engineers evaluating coding models, you should critically assess performance claims, especially when cost savings are significant. While Kimi K2.7 Code offers a compelling price point at \$4.00 per million output tokens, its reliance on unverified, in-house benchmarks and underperformance against competitors on those same metrics suggests caution. Prioritize models with transparent, publicly verifiable benchmarks to ensure you are not compromising code quality for cost efficiency.

Key insights

Kimi K2.7 Code offers extreme cost efficiency but lacks independent performance verification, raising trust concerns.

Principles

• Cost-efficiency can mask performance gaps.
• Proprietary benchmarks hinder trust.
• Open-weight models challenge market pricing.

In practice

• Evaluate models beyond stated costs.
• Prioritize models with public benchmarks.
• Scrutinize vendor-specific performance claims.

Topics

• Kimi K2.7 Code
• Code Generation
• Model Benchmarking
• Cost Efficiency
• Open-weight Models
• Mixture-of-Experts

Best for: NLP Engineer, CTO, VP of Engineering/Data, AI Engineer, Machine Learning Engineer, AI Scientist

Related on AIssential

• Open model Kimi K2.7 Code undercuts GPT-5.5 and Claude by up to 12x on price per token — Kimi K2.7 Code offers a cost-effective, agent-optimized open-source model for complex coding, despite trailing top benchmarks.
• Kimi K2.7-Code cuts thinking tokens 30% — but practitioners say the benchmarks don't check out — Kimi K2.7-Code offers claimed token efficiency and proprietary benchmark gains, but independent tests reveal mixed performance and raise benchmark transparency concerns.
• Kimi K2.7 Code vs Claude Mythos & Fable — Kimi K2.7 Code offers cost-effective, open-weight agentic coding, challenging Claude's performance lead in specific workflows.
• The Dragon’s Code vs The Anthropic Giant: How Kimi K2.5, — New Chinese open-source models are challenging Claude Opus 4.6's dominance in AI coding agents.
• Kimi K2.6 Shipped. Palantir Published. The West Is Walking Backwards. — Chinese open-weight models are achieving frontier capabilities and cost-effectiveness, challenging Western closed-API dominance.
• 5 Tips to Cut Claude Code Token Usage by 30% — Optimizing Claude Code usage through structured context and model selection significantly reduces token costs.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Towards AI - Medium.