Kimi

What it is for

Kimi is a signed-out Moonshot assistant: the official Help still (see the cover: a real thread — summarize an Alphabet earnings URL, Fetch URLs, then the Q1 call summary) is the first-run canvas. Use it when the job is talk to Kimi (or call the API), not when you want a local model.

The chat product is not open source. The API is pay-as-you-go per 1M tokens — the pricing doc says “no subscription plan.” Chat membership and Kimi Code are named as different products on that page; the table prints token rows only.

What makes it unique: the cover is Kimi’s own Help Center conversation (search / fetch a URL), not an empty composer.

Key features

• Signed-out composer — “Ask anything, or ask an agent…”, Instant High (cover).
• Agent chips — Slides, Deep Research, Docs, Build, Sheets, Design.
• Sidebar — New chat, My kimi, Plugins, Scheduled, Projects.
• Kimi API — PAYG tokens; cache write / cached input / input / output columns.
• K3 long context — 1,048,576 tokens on the K3 row we copied.

Who it's for

Best fit: people who want a signed-out Kimi window, or developers who will pay per 1M on the API.

Poor fit: people who need a local runner (Ollama) or a multi-vendor shelf (Poe).

Pricing

Verified 2026-10-21 (Asia/Shanghai) from platform.kimi.ai/docs/pricing/chat. Units are per 1M tokens. Re-check before you buy. Chat membership / Kimi Code are named as different products — no membership $ was on this table.

• kimi-k3 — cache write $3.00 (5 min TTL) / $6.00 (1 h TTL); cached input $0.30; input $3.00; output $15.00. Context 1,048,576.
• kimi-k2.7-code — input cache-hit $0.19 / miss $0.95; output $4.00. Context 262,144.
• kimi-k2.7-code-highspeed — cache-hit $0.38 / miss $1.90; output $8.00. Context 262,144.
• kimi-k2.6 — cache-hit $0.16 / miss $0.95; output $4.00. Context 262,144.

File extract/storage APIs were temporarily free on that page. Prices exclude tax.

Limitations / costs

• No membership dollar on the API table. Do not invent a Kimi Plus seat.
• $15/1M output is K3. Code models are cheaper and shorter-context.
• Chat and API are different SKUs. The composer does not print these rows.
• Not local. Prompts leave your machine.

Alternatives

• ChatGPT — hosted general assistant; Plus $20/month on Help.
• DeepSeek — chat + API with public token rows.
• Poe — multi-bot shelf with a public subscription card.

Official tutorial

Official Kimi AI channel video (channel linked from platform.kimi.ai): Get Started with Kimi Work - Tutorial #3 (oEmbed author: Kimi AI).

Written Getting started with Kimi plus Model Inference Pricing.

Website

https://www.kimi.com

Source

Catalog row curated for AI Tools Nav. Cover is the official Help Center search conversation (Alphabet Q1 earnings summary). API dollars checked on platform.kimi.ai/docs/pricing/chat. No membership $ on that page. Tutorial: Get Started with Kimi Work on the official Kimi AI channel (oEmbed).

Last verified

2026-10-21 (Asia/Shanghai) editorial check. K3 $3 / $15 per 1M in/out; K2.7-code $0.19 / $0.95 / $4.00. Membership unnamed in dollars.