Kimi 月之暗面
Long-context chat assistant from Moonshot AI
Overview
Moonshot AI's chat assistant, built on the open-weight Kimi K-series — K2.6 and the coding-focused K2.7 (2026), 1T-parameter MoE models with a 256K-token context. Strong at long-document Q&A and multilingual chat, with an aggressively priced API.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 11 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- 6 agent credits
- 200 professional database calls
- 1 concurrent task
- $15/mo billed annually
- 60 agent credits, Deep Research
- Kimi Code, Slides and Websites tools
- $31/mo billed annually
- 150 agent credits, 5x Kimi Code credits
- Kimi Claw access
- $79/mo billed annually
- 360 agent credits, 15x Kimi Code credits
- 120 Agent Swarm uses
- $159/mo billed annually
- 720 agent credits, 30x Kimi Code credits
- 240 Agent Swarm uses
- K2.5: $0.60/$3.00 per M in/out
- K2.6: $0.95/$4.00 per M in/out
- Cached input from $0.10/M
Use cases
What you can produce with Kimi 月之暗面
- Analyse and synthesise a 500+ page legal or technical document in a single pass — feeding an entire corporate contract, regulatory filing, or codebase into the 256K context window and receiving a structured summary with cross-references
- Run an autonomous overnight coding project — orchestrating up to 300 sub-agents to refactor a multi-file codebase, write unit tests, and produce a pull-ready diff without human intervention
- Generate a production-ready website with responsive design from a single descriptive prompt using Kimi Work's web-building agent and native vision capabilities
- Produce a multi-source competitive intelligence report — dispatching agent swarms to search, scrape, and synthesise findings from dozens of URLs into a formatted research document in 2-3 minutes
- Create a structured presentation deck from raw research notes, with Kimi's Slides tool generating a genuinely researched first draft including section flow and supporting data
- Build and integrate MCP tool-use agent pipelines at significantly lower cost (~80-88% cheaper than Claude/OpenAI equivalents) using the open-weight K2.7 Code model as a drop-in replacement in Cline, RooCode, or similar frameworks
- Extract structured insights from video or image inputs — using native multimodal vision to analyse charts, diagrams, UI screenshots, or video walkthroughs and produce actionable written output
ASEAN Perspective
Kimi 月之暗面 in Southeast Asia
中国市场最受关注的长上下文 AI 助手之一。在企业 RAG 与研究类工作场景中与 DeepSeek、文心一言、豆包形成直接竞争。
Kimi is Moonshot AI's consumer front-end for the open-weight K-series models, and in 2026 it's no longer a China-only story: kimi.com, the Kimi Code CLI and an English API platform give it a real international footprint, backed by K2.6 and K2.7-Code with 256K-token context. Long-document Q&A, research synthesis and Chinese-language work remain its strongest suits, and API pricing dramatically undercuts Western rivals. The caveats: outputs run verbose, the newest coding model forces thinking mode on, launch benchmarks are all vendor-published, and Moonshot's Chinese ownership raises data-governance questions an ASEAN business must answer for itself. As a free long-context assistant or a cheap API workhorse it earns a place on the shortlist; as a default enterprise assistant with clear data terms, the global leaders still have the edge.
What people say
$0.95 per million input tokens, $4.00 out — that's the pricing that keeps pulling developers toward Moonshot AI's Kimi, roughly 75-90% below OpenAI and Anthropic equivalents for comparable work. The consumer chat app rides on the open-weight K-series: K2.6 landed April 20, 2026 (1T-parameter MoE, 32B active, 256K context, agent-swarm orchestration up to 300 sub-agents), and K2.7-Code followed June 12 with a claimed 21.8% jump on Moonshot's own coding bench and about 30% less reasoning-token burn. Both ship on Hugging Face under a modified MIT licence; Kimi Code membership plans start at $19/month.
Reddit sentiment is measured rather than breathless. Consistent praise: the free web chat dropped its daily message cap in early 2026, long-context handling genuinely digests full documents without truncation, and coding output competes with frontier Western models. Consistent gripes: verbose answers, K2.7-Code's thinking mode is forced on (you can't disable it, and the reasoning adds real latency), and privacy questions recur because Moonshot is a Chinese company with Alibaba investment. Self-hosting the open weights is theoretical for most — K2.5 alone weighed 595GB, and Moonshot's own Reddit AMA filled up with requests for smaller coder variants.
One caution on the spec sheet: every benchmark published for K2.7 at launch was Moonshot's own. No independent SWE-bench Verified or LiveCodeBench numbers existed at release, so treat the coding claims as vendor-reported until third parties weigh in. For long-document Q&A and cheap API workloads it's a serious tool; for teams that need clear non-China data terms, that's still the sticking point.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including Kimi 月之暗面's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Kimi 月之暗面 unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Kimi 月之暗面 directly →
Spotted something out of date? Suggest an update →
Alternatives to Kimi 月之暗面
More in LLMs & Chat
