Breaking

Moonshot AI Publishes Official Cheat Sheet for Kimi Code CLI

June 10, 2026 at 10:01 EDT

Software Dev & Coding
AI Agents
Foundation Models

Moonshot AI has released an official cheat sheet compiling the commands and shortcuts for its terminal-based AI coding agent, Kimi Code CLI. Provided as a continuously updated resource on a dedicated landing page, the Kimi Code Cheat Sheet lists the tool's key operations in one place.

June 10, 2026 · Moonshot AI

Kimi Code CLI: an open, low-cost terminal coding agent takes aim at Claude Code

Moonshot ships an official cheat sheet for its Apache-2.0 coding agent, powered by the 1-trillion-parameter K2.6 model — pitched as a cheaper, open migration target from Cursor and Claude Code.

8×+

cheaper input pricing than Claude

262K

token context window in K2.6

6,400+

GitHub stars

88%

coding cost cut reported vs Claude

SWE-Bench Pro: K2.6 edges the frontier

Vendor-reported scores — higher is better.

58.6%

Kimi K2.6

57.7%

GPT-5.4

53.4%

Claude Opus 4.6

K2.6 benchmark scorecard

BrowseComp (Agent Swarm)86.3%

LiveCodeBench v689.6%

SWE-Bench Verified80.2%

Terminal-Bench 2.066.7%

Humanity's Last Exam (tools)54.0%

Terminal-Bench 2.0 reportedly up 15.9 points over K2.5; Agent Swarm scaled from 100 → 300 sub-agents.

What developers praise

Open-source (Apache 2.0), low-cost alternative
13-hour autonomous run refactoring a 4,000+ line financial engine, 1,000+ tool calls
~88% cost reduction reported vs Claude
Multimodal: code from dragged-and-dropped video

Caveats noted

Trails Claude Opus 4.7 on some tasks — lease/streaming: 68 vs 91 / 100
Benchmarks are vendor-leaning; need independent verification
Weekly query limits (~2,048/wk at $20 tier)
Agent Swarm stability is model-dependent

Official API pricing — K2.6

$0.60/M

input tokens

$2.50–3.40/M

output tokens

60–80%

extra reduction on cache hits

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.