Kimi
Kimi is Moonshot AI's consumer assistant and API surface. The live site at kimi.com now sells "Kimi AI with K3" for agentic coding and knowledge work, not the 256K / K2-only product this page used to describe. The model card for the consumer and API flagship is Kimi K3. The terminal coding agent is Kimi Code CLI. This page is the product, not a single checkpoint.
API traffic goes to https://api.moonshot.ai/v1 with MOONSHOT_API_KEY. Docs and keys live on platform.kimi.ai. Rechecked 2026-08-17: the chat catalog leads with kimi-k3, plus kimi-k2.7-code-highspeed for coding speed and kimi-k2.6 as a prior generation.
Key Features
- K3 as the default brain: Flagship chat and API model. It always reasons. Official
reasoning_effortvalues arelow,high, andmax(defaultmax). - 1,048,576-token API context: K3's documented context is 1M tokens, not the old 256K window.
- Coding siblings: Use Kimi Code in the terminal, or call
kimi-k2.7-code-highspeedwhen you want a faster coding model instead of full K3. - Consumer + API split: kimi.com is the assistant. platform.kimi.ai is the developer console.
Use Cases
- Long briefs and repos that need a 1M-token window without you chunking the source first.
- Agentic coding when you want Moonshot's own assistant rather than a third-party wrapper.
- Teams already on Kimi Code who still need a browser chat for specs, reviews, and research.
Limitation: membership copy on kimi.com says Kimi and Kimi Code benefits will split. Do not assume one plan covers both products forever. English-first teams should also test latency and language quality themselves. We did not re-verify China-only student CNY plans.
Pricing
Consumer membership on kimi.com/membership/pricing (2026-08-17). Annual effective prices are listed first; monthly list price is in parentheses.
| Plan | Price | Notes |
|---|---|---|
| Adagio | $0 | Free consumer tier |
| Moderato | $15 / year effective ($19 monthly) | Paid starter |
| Allegretto | $31 / year effective ($39 monthly) | Mid paid |
| Allegro | $79 / year effective ($99 monthly) | Up to 1M chat context |
| Vivace | $159 / year effective ($199 monthly) | Top published consumer plan |
K3 API on platform.kimi.ai/docs/pricing/chat-k3: $3.00 input miss, $0.30 cache hit, $15.00 output, per 1M tokens. Context 1,048,576. We dropped older $0.60 / $2.50 API figures, third-party reseller prices, a 36M MAU claim, and a $4.3B valuation. monthlyVisits: 3600000 is the existing catalog figure, not a new audit.
Getting Started
- Open kimi.com for chat, or platform.kimi.ai for an API key.
- Set
MOONSHOT_API_KEYandbase_url=https://api.moonshot.ai/v1. - Call
kimi-k3unless you have a reason to pinkimi-k2.7-code-highspeedorkimi-k2.6. - If the job is a repo, start in Kimi Code CLI instead of pasting files into chat.
Frequently Asked Questions
Is Kimi still a 256K chatbot?
No. The current flagship API context is 1,048,576 tokens on K3.
Is Kimi the same as Kimi K3 or Kimi Code?
No. Kimi is the assistant and API product. Kimi K3 is the model. Kimi Code is the terminal agent.
Can I keep using K2.6?
Yes. The platform still lists kimi-k2.6 and a K2.7 coding speed model next to K3.
Alternatives
- Doubao: ByteDance consumer assistant if you are already in that stack.
- ChatGPT: OpenAI's consumer chat if you want that ecosystem.
- Kimi Code CLI: Terminal-first coding instead of the website.
Tips
- Read K3 pricing on the platform, not old blog posts that still quote $0.60 input.
- Treat membership and API bills as separate. The music-named plans are consumer seats.
- Recheck the membership banner. Kimi vs Kimi Code entitlements are being split.
Conclusion
Kimi is Moonshot's live assistant, now led by K3 with a 1M-token API and a small family of K2.x coding models. Start at kimi.com or the K3 pricing page, then jump to the K3 model page if you need weights and architecture.
Comments
No comments yet. Be the first to comment!
Related Tools
Kimi K3
www.kimi.com
Moonshot AI's open-weight 2.8T multimodal agentic model with 1M-token context, the world's first open 3T-class model rivaling closed frontier models.
Kimi Code CLI
www.kimi.com/code
Moonshot terminal agent. Rechecked: CLI is free; consumer Moderato $19/month or $15/year. K2.7 Code API is CNY, not unlimited $19 K2.
Doubao
www.doubao.com
ByteDance's consumer AI assistant. Chat, writing, coding, and Seedance video in one app. Official homepage currently offers Seedance 2.0 for free after login.
Related Insights
Hook OpenCode Go into Codex on Windows. Do Not Open a Second Toolkit.
A ChatGPT-signed Codex desktop app still shows mostly GPT in the picker. On Windows, enable only OpenCode Go and the Grok, GLM, Kimi, DeepSeek, and MiniMax models you already pay for appear in the same selector. Keys stay local. Native GPT stays put.
Stop Cramming AI Assistants into Chat Boxes: Clawdbot Picked the Wrong Battlefield
Clawdbot is convenient, but putting it inside Slack or Discord was the wrong design choice from day one. Chat tools are not for operating tasks, and AI isn't for chatting.
Seven AI Coding CLIs, Six Months: No Matter How Strong the Model, Work Needs Supervision
Claude Code, Codex, opencode, pi, omp and DeepSeek Harness all have personalities. After six months of deep use I run a division of labor: pi for the fastest cheapest reviews, omp for complex PRs, DeepSeek Harness on V4 Flash for high-frequency low-cost review, and Claude Code, Qoder and Cursor for writing. No matter how strong the model, work needs supervision — ideally from an independent third party.