# Kimi K2.7 Code ![Organization](https://img.shields.io/badge/Org-Moonshot-2d9de8) ![License](https://img.shields.io/badge/License-Proprietary-d29922) ![Released](https://img.shields.io/badge/Released-Jul%202026-555) ![Focus](https://img.shields.io/badge/Focus-Coding-1f6feb) > **Quick answer:** Kimi K2.7 Code is Moonshot AI's July 2026 coding-specialised model in the Kimi K2 family. It scores 36.7% on LHTB. As a code-focused variant of Kimi K2.7, it is designed for coding agent tasks and software engineering. ## At a Glance **Where Kimi K2.7 Code leads** - Coding specialisation within Kimi K2 family - March 2026 knowledge cutoff - Recent release in the Kimi K2 ecosystem **Where it lags** - 36.7% LHTB — moderate long-horizon task performance - Limited public benchmark coverage available - Proprietary API-only **Best for:** Kimi K2 ecosystem coding tasks; teams using Moonshot AI services for software engineering. ## What Kimi K2.7 Code Is Kimi K2.7 Code is the coding-specialised variant in Moonshot AI's Kimi K2 model family. The K2 family (including K2 Instruct with 59.1% BFCL) has been Moonshot's primary open-weight offering; K2.7 Code appears to be a later closed-weights proprietary variant specifically tuned for coding tasks. Limited benchmark data is publicly available for this model beyond the LHTB score. Teams considering Kimi K2 for coding should compare against the open-weight Kimi K2 Instruct (Apache 2.0) which offers full self-hosting capability. ## Specifications | Field | Value | |---|---| | Organization | Moonshot AI | | License | Proprietary (API only) | | Release date | July 2026 | | Knowledge cutoff | March 2026 | | Modality | Text / Code | ## Pricing Available via Moonshot AI API (moonshot.cn). Refer to current Moonshot pricing. ## Public Benchmark Scores | Benchmark | Score | Source | Date | |---|---|---|---| | [LHTB](/benchmarks/lhtb) | 36.7% | Benchgen evaluation | 2026-07 | ## Kimi K2.7 Code vs Alternatives | Model | LHTB | License | Type | |---|---|---|---| | **Kimi K2.7 Code** | 36.7% | Proprietary | Coding | | [Kimi K2 Instruct](/models/moonshot/kimi-k2-instruct) | — | Apache 2.0 | General | | [HY3](/models/hungyuan/hy3) | 28.8% | Proprietary | General | Kimi K2 Instruct (Apache 2.0) is the recommended choice for teams requiring open weights. Kimi K2.7 Code is for teams integrated into the Moonshot AI API ecosystem. ## Related Models & Benchmarks - Compare: [Kimi K2 Instruct](/models/moonshot/kimi-k2-instruct) · [HY3](/models/hungyuan/hy3) · [GPT-5 Codex](/models/openai/gpt-5-codex) - Benchmarks: [LHTB](/benchmarks/lhtb) ## Frequently Asked Questions

What is Kimi K2.7 Code?

Moonshot AI's July 2026 coding-specialised model scoring 36.7% LHTB. A proprietary coding variant of the Kimi K2 family.

Should I use Kimi K2.7 Code or Kimi K2 Instruct?

Kimi K2 Instruct (Apache 2.0) for open-weight flexibility. Kimi K2.7 Code for API-only use cases within the Moonshot ecosystem.

--- *Specs from Moonshot AI's Kimi K2.7 Code release (July 2026) and Benchgen evaluations. Last updated 2026-07-24.*

Kimi K2.7 Code — Training jobs

Training jobs