The Lorphic guide provides a consolidated overview of the Kimi K2 model family — K2, K2.5, K2.6, and K2.7 Code — with architecture, version-by-version differences, benchmark data with proper attribution, API setup, and practical positioning. Every model in the K2 family shares a 1-trillion-parameter MoE foundation with The guide emphasizes that MoE architecture is the reason Kimi models can be priced competitively against dense models of comparable scale, since inference cost scales with active parameters rather than total parameters. By walking through the naming and capability differences across versions, it helps developers select

Kimi K2.7 Code, Moonshot AI's open-weight coding-specialised model, appeared on Hugging Face on June 12, 2026, under a Modified MIT license. The model is a 1-trillion-parameter Mixture-of-Experts transformer with 32 billion active parameters per token, 384 experts (8 selected + 1 shared), 61 layers (1 dense), Multi-hea The guide highlights a roughly 30% reduction in reasoning-token usage compared to Kimi K2.6, which directly lowers the cost of agentic coding loops, alongside higher scores on Moonshot's coding benchmarks. At the time of writing, no independent benchmarks (SWE-bench, GPQA) had been published, and Moonshot had not yet i

Kimi K2.7-Code was released by Moonshot AI on June 12, 2026 as an open-source, coding-focused successor to Kimi K2.6, with weights published on Hugging Face under a Modified MIT license. The release is paired with distribution: immediate availability through the Kimi API at platform.moonshot.ai and through Kimi Code, M Moonshot reports three benchmark gains over K2.6: +21.8% on Kimi Code Bench v2 (50.9 → 62.0), +11.0% on Program Bench, and +31.5% on MLS Bench Lite, alongside approximately 30% lower reasoning-token usage. The article explicitly flags these as first-party vendor self-reports with no independent verification at launch.