The TechTimes article dated June 15, 2026 reports that Moonshot AI released Kimi K2.7-Code on June 12, 2026 and on June 15 announced a HighSpeed Mode delivering up to six times faster token throughput. The HighSpeed variant promises around 180 tokens per second on median-length coding inputs and up to 260 tokens per se The article notes that Moonshot has not submitted the model to any independent coding benchmark, leaving developers with vendor-reported numbers as their only purchasing guide. The ~30% reasoning-token efficiency gain versus K2.6 rests entirely on five proprietary benchmarks Moonshot designed and ran itself. K2.7-Code

The LLM Reference page for Kimi K2.7-Code HighSpeed, last refreshed June 22, 2026, lists it as a serving variant of Kimi K2.7-Code optimized for throughput, released June 15, 2026. It documents a 1T-parameter MoE architecture with 32B active parameters, 384 experts with 8 selected per token, a MoonViT vision encoder, a The page lists one tracked provider route with input priced at $1.90 per 1M tokens, output at $8.00 per 1M tokens, and cache reads at $0.380. It positions the HighSpeed variant as best suited for interactive or latency-bound workflows, with the standard variant preferred for correctness-sensitive long-horizon agentic w

The MarkTechPost coverage dated June 13, 2026 reports on Moonshot AI's release of Kimi K2.7-Code as a coding-focused, agentic model targeting long-horizon software engineering rather than general chat. The model weights ship on Hugging Face under a Modified MIT license and are also accessible through the Kimi API and K The article echoes Moonshot's benchmark claim of +21.8% on Kimi Code Bench v2 over K2.6 as its headline performance number. As an aggregator/editorial site, it serves as background context for the K2.7-Code release that underpins the Highspeed tier, rather than as a primary technical reference.

The Digital Applied blog post dated June 12, 2026 covers Moonshot AI's release of Kimi K2.7-Code, the open-source coding model underlying the Highspeed tier, with weights live on Hugging Face under a Modified MIT license. Moonshot reports benchmark gains of +21.8% on Kimi Code Bench v2 (50.9 to 62.0 versus K2.6), +11.0 The release pairs the model with Kimi Code, Moonshot's terminal-first coding agent, with membership plans listed from $19 per month, positioning the launch as a model-plus-platform play. API access is available through platform.moonshot.ai, and the strategic read is that Moonshot is shipping both weights and a subscrip